USF Libraries
USF Digital Collections

Strain, personality traits, and deviance among adolescents

MISSING IMAGE

Material Information

Title:
Strain, personality traits, and deviance among adolescents moderating factors
Physical Description:
Book
Language:
English
Creator:
Wareham, Jennifer J
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
strain -- psychopathy -- delinquency -- drug use -- social control -- delinquent peers
Psychopathy
Delinquency
Drug use
Social control
Delinquent peers
Dissertations, Academic -- Criminology -- Doctoral -- USF
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: General strain theory has received a fair amount of empirical support and theoretical elaboration over the past several years. Since the introduction of general strain theory, Agnew and others have attempted to increase the comprehensiveness of the processes involved in strain theory. Until recently, the general strain theory literature has ignored what Agnew and associates (Agnew, Brezina, Wright, & Cullen, 2002) argue may be one of the most important conditioning effects of the strain-crime relationship, namely the dispositions or personality traits of the individual experiencing strain. Recently, Agnew and associates (2002) published results from a study examining the conditioning effects of personality traits (i.e., negative emotionality and low constraint) on the strain-delinquency relationship. Their findings indicated that certain personality traits significantly condition the effect of strain on delinquency. Research has suggested that more severe personality ^and behavioral traits, such as psychopathy, also influence criminality. The present study examined moderating effects of both personality dispositions and psychopathic behavioral features among a sample of 137 youths referred to juvenile diversion by the court system. The results suggest that personality dispositions and psychopathic behavioral features do not significantly moderate the strain-delinquency relationship. In addition, this study conducted ad hoc analyses examining whether or not delinquency significantly increases the likelihood that subsequent strain and delinquency will result (i.e., a state dependence explanation (see Nagin & Farrington, 1992; Nagin & Paternoster, 1991)). Moderating effects of personality and psychopathy were also included in this model. Further, the role of strain as a mediator for the personality and psychopathy link to delinquency was tested. The findings suggest that delinquency exacerbated subsequent strain and delinquency levels among thes e youths. Personality and psychopathic features did not moderate the strain-delinquency relationship. Strain did not significantly moderate the personality-delinquency relationship. Limitations and implications for future research and policy are discussed.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2005.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Jennifer J. Wareham.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 261 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001911252
oclc - 173683502
usfldc doi - E14-SFE0001212
usfldc handle - e14.1212
System ID:
SFS0025533:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001911252
003 fts
005 20071005101040.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 071005s2005 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001212
040
FHM
c FHM
035
(OCoLC)173683502
049
FHMM
090
HV6025 (ONLINE)
1 100
Wareham, Jennifer J.
0 245
Strain, personality traits, and deviance among adolescents :
b moderating factors
h [electronic resource] /
by Jennifer J. Wareham.
260
[Tampa, Fla] :
University of South Florida,
2005.
3 520
ABSTRACT: General strain theory has received a fair amount of empirical support and theoretical elaboration over the past several years. Since the introduction of general strain theory, Agnew and others have attempted to increase the comprehensiveness of the processes involved in strain theory. Until recently, the general strain theory literature has ignored what Agnew and associates (Agnew, Brezina, Wright, & Cullen, 2002) argue may be one of the most important conditioning effects of the strain-crime relationship, namely the dispositions or personality traits of the individual experiencing strain. Recently, Agnew and associates (2002) published results from a study examining the conditioning effects of personality traits (i.e., negative emotionality and low constraint) on the strain-delinquency relationship. Their findings indicated that certain personality traits significantly condition the effect of strain on delinquency. Research has suggested that more severe personality ^and behavioral traits, such as psychopathy, also influence criminality. The present study examined moderating effects of both personality dispositions and psychopathic behavioral features among a sample of 137 youths referred to juvenile diversion by the court system. The results suggest that personality dispositions and psychopathic behavioral features do not significantly moderate the strain-delinquency relationship. In addition, this study conducted ad hoc analyses examining whether or not delinquency significantly increases the likelihood that subsequent strain and delinquency will result (i.e., a state dependence explanation (see Nagin & Farrington, 1992; Nagin & Paternoster, 1991)). Moderating effects of personality and psychopathy were also included in this model. Further, the role of strain as a mediator for the personality and psychopathy link to delinquency was tested. The findings suggest that delinquency exacerbated subsequent strain and delinquency levels among thes e youths. Personality and psychopathic features did not moderate the strain-delinquency relationship. Strain did not significantly moderate the personality-delinquency relationship. Limitations and implications for future research and policy are discussed.
502
Dissertation (Ph.D.)--University of South Florida, 2005.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 261 pages.
Includes vita.
590
Adviser: Richard Dembo, Ph.D.
653
strain
psychopathy
delinquency
drug use
social control
delinquent peers
Psychopathy.
Delinquency.
Drug use.
Social control.
Delinquent peers.
690
Dissertations, Academic
z USF
x Criminology
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1212



PAGE 1

Strain, Personality Traits, and Devian ce among Adolescents: Moderating Factors by Jennifer J. Wareham A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Criminology College of Arts and Sciences University of South Florida Co-Major Professor: Richard Dembo, Ph.D. Co-Major Professor: John K. Cochran, Ph.D. Christine S. Sellers, Ph.D. Norman Poythress, Ph.D. Date of Approval: July 13, 2005 Keywords: strain, psychopathy, delinquency, dr ug use, social control, delinquent peers Copyright 2005, Jennifer J. Wareham

PAGE 2

Dedication This dissertation is dedicated to my daughter, Jadziah, who was born the semester I was preparing for my Comprehensive Exams, and Kristen, my best friend, soul mate, and advocate during times of need. I have b een blessed to find the support and love in you during my adulthood that was always lacking in my youth.

PAGE 3

Acknowledgments I wish to express my deepest appreciation to my Committee members and the faculty and staff of the Department of Crimi nology at the University of South Florida. There are several people in particular that I w ould like to take a mome nt to thank. To Dr. Richard Dembo, I have learned so much while under your tutelage over the fast four years. You are an amazing academe and humanitarian. Thank you for your kindness and patience. You have inspired me to become a better human being. To Dr. John Cochran, thank you for all of the advice and time you de dicated to my learning and growth as an academic over the past decade. I do not know what I would have done this year without you. To Dr. Christine Sellers, thank you for help ing me to become a better writer, critical thinker, and criminologist. I a ppreciate all of your advice and friendship this past year. I would have been lost without you. For me, th e journey to this moment has often seemed long, labor-intensive, and exhausting, tinged by a seed of doubt. At last, I am at peace with my decision to be a criminologist. I ha ve no regrets, and only hope that I can make everyone that helped me along the way proud.

PAGE 4

i Table of Contents List of Tables................................................................................................................. ....iv List of Figures................................................................................................................ ....vi Abstract....................................................................................................................... ......vii Chapter 1 Introduction......................................................................................................... 1 Chapter 2 General Strain Theory.......................................................................................13 Anomie and Classic Strain Theory........................................................................13 Criticism of Classic Strain Theory.........................................................................18 General Strain Theory............................................................................................22 Types of Strain...........................................................................................23 Negative Affect..........................................................................................32 Coping Mechanisms...................................................................................33 Other Conditioning Factors........................................................................34 Empirical Support for General Strain Theory............................................40 Strain-delinquency.........................................................................40 Mediating influence of negative affect..........................................41 Moderation/mediation of coping mechanisms and other factors.......................................................................43 Distinguishing General Strain Th eory from Social Control and Social Learning Theories...........................................................................45 Social Control Theory................................................................................47 Social Learning/Differential Association Theory......................................49 Two Weaknesses of Tests of GS T: Tautology and Falsifiability..............51 Chapter 3 Personality Characte ristics Affecting Deviance................................................58 Trait or Motive?.....................................................................................................59 Personality Traits...................................................................................................61 Temperament.............................................................................................63 Hierarchy of Personality............................................................................64 Personality Traits Affecting Deviance.......................................................66 Psychopathic Features Affecting Deviance...........................................................69 Psychopathy: Taxon or Dimension?..........................................................74 Juvenile Psychopathy.................................................................................77

PAGE 5

ii The Tautology of Personality Tr aits and Psychopathic Features and Crime...................................................................................................83 Chapter 4 General Strain Theory a nd Personality: Another Theoretical Elaboration.............................................................................................................89 GST and the Conditioning Effects of Personality and Psychopathic Features................................................................................90 The First Test of GST and Personality..................................................................92 The Proposed Study...............................................................................................95 GST, Personality, Delinquency, and Drug Use Problems.......................100 Chapter 5 Method............................................................................................................110 Sample..................................................................................................................111 The Arbitration Intervention Workers Service (AIW).............................111 Sociodemographic Information at Time 1...............................................114 Measures..............................................................................................................117 The Comprehensive Adolescen t Severity Inventory (CASI)...................118 Deriving Appropriate Measures from the CASI......................................120 Strain Measures........................................................................................128 Family disruption.........................................................................128 Family abuse/neglect...................................................................129 Peer strain.....................................................................................130 Social Control Measures..........................................................................130 Low parental attachment/commitment........................................130 Low school attachment................................................................131 Low school commitment..............................................................131 Social Learning/Differen tial Association Measures................................132 Personality and Psychopathic Features....................................................133 APSD psychopathic features........................................................134 YPI psychopathic features...........................................................136 CASI mental health measures......................................................137 Delinquency.............................................................................................139 Drug and Alcohol Use Problems.............................................................141 Drug problems.............................................................................144 Description of Observed Variables..........................................................147 Chapter 6 Results.............................................................................................................1 49 Analytic Strategy.................................................................................................149 Findings................................................................................................................151 Initial GST Models..................................................................................151 SEM of GST for Time 1 Only.................................................................155 Variable Adjustment................................................................................156 Supplemental Analyses........................................................................................165 Path Analyses of Strain Leading to Delinquency/Drugs.........................165 Path Analyses of Delinque ncy/Drugs Leading to Strain.........................168

PAGE 6

iii Chapter 7 Discussion.......................................................................................................196 Limitations...........................................................................................................203 Implications..........................................................................................................206 References..................................................................................................................... ...211 Appendices..................................................................................................................... ..252 Appendix A: Varimax Rotated Explor atory Factor Analyses Results For Family, Peer, and School Items during Twelve Months Prior to Baseline Interview (N = 137).....................................................253 Appendix B: Zero-Order Correla tion Matrix for Final Measures.......................256 About the Author...................................................................................................End Page

PAGE 7

iv List of Tables Table 1: Sociodemographic Informati on at Time of Baseline Interview (N = 137)..............................................................................................................116 Table 2: CFA Standardized Loadings for Family Items for Time 1 and Time 2..................................................................................................................124 Table 3: CFA Standardized Loadings fo r Peer Items for Time 1 and Time 2.................126 Table 4: CFA Standardized Loadings for School Items for Time 1 and Time 2..................................................................................................................127 Table 5: Self-Reported Delinquency (N = 137)...............................................................141 Table 6: CFA Standardized Loadings for Drug Use Problems for Time 1 and Time 2...........................................................................................................145 Table 7: Descriptive Statisti cs for Observed Measures...................................................146 Table 8: Delinquency (log) on Strain, So cial Control, and Delinquent Peer Factors Estimates (Standardized Estimates)........................................................153 Table 9: Drug Problems Factor on Stra in, Social Control, and Delinquent Peer Factors Estimates (Standardized Estimates)................................................154 Table 10: Delinquency (log) and Drug Problems Factor on Time 1 Strain Estimates (Standard ized Estimates).....................................................................156 Table 11: Descriptive Statistics for Adjusted Delinquency and Drug Measures..............................................................................................................158 Table 12: Recoded Delinquency and Drug Use on Strain Estimates (Standardized Estimates).....................................................................................160 Table 13: Recoded Delinquency and Dr ug Use on Strain (Time 1 Only) Estimates (Standard ized Estimates).....................................................................161

PAGE 8

v Table 14: Descriptive Statistics for St rain, Social Control, and Social Learning Indexes..................................................................................................162 Table 15: Delinquency (log) and Drug Problems Factor on Time 1 Strain Index Estimates (Standardized Estimates)...........................................................165 Table 16: Unstandardized Estimates for Path Analyses of Self-Reported Delinquency and Drug Problem s-Usage on Strain (T1)......................................167 Table 17: Unstandardized Parameter Estimates of the Path Models of Delinquency (log), Strain (T2), a nd Personality Characteristics (T1) (N = 137)......................................................................................................176

PAGE 9

vi List of Figures Figure 1: A Model of Ge neral Strain Theory.....................................................................39 Figure 2: Non-Recursive Model of St rain and Personality Features on Delinquency.........................................................................................................106 Figure 3: Non-Recursive Model of St rain and Personality Features on Drug Use Problems..............................................................................................107 Figure 4: Ad Hoc Contemporaneous Mode l of Strain, Social Control, and Delinquent Peers..................................................................................................172

PAGE 10

vii Strain, Personality Traits, and Devian ce among Adolescents: Moderating Factors Jennifer J. Wareham ABSTRACT General strain theory has received a fair amount of empirical support and theoretical elaboration over the past severa l years. Since the in troduction of general strain theory, Agnew and others have attempted to increase the comprehensiveness of the processes involved in strain theory. Until recen tly, the general strain theory literature has ignored what Agnew and a ssociates (Agnew, Brezina, Wright, & Cullen, 2002) argue may be one of the most important conditioning effects of the strain-crime relationship, namely the dispositions or personality tra its of the individual e xperiencing strain. Recently, Agnew and associates (2002) published results from a study examining the conditioning effects of personal ity traits (i.e., negative emo tionality and low constraint) on the strain-delinquency relationship. Their findings indi cated that certain personality traits significantly condition the effect of strain on delinque ncy. Research has suggested that more severe personality and behavioral traits, such as psychopathy, also influence criminality. The present study examined moderating effects of both personality dispositions and psychopathic behavioral features among a sample of 137 youths referred to juvenile diversion by the court system. The results suggest that personality dispositions and

PAGE 11

viii psychopathic behavioral featur es do not significantly mode rate the strain-delinquency relationship. In addition, th is study conducted ad hoc analyses examining whether or not delinquency significantly increase s the likelihood that subsequent strain and delinquency will result (i.e., a state dependence explan ation (see Nagin & Farrington, 1992; Nagin & Paternoster, 1991)). Moderating effects of personality and psychopathy were also included in this model. Further, the role of strain as a mediator for the personality and psychopathy link to delinquency was tested. The findings suggest that delinquency exacerbated subsequent strain and delinquenc y levels among these youths. Personality and psychopathic features did not moderate th e strain-delinquency re lationship. Strain did not significantly moderate the personality -delinquency relationship. Limitations and implications for future research and policy are discussed.

PAGE 12

1 Chapter 1 Introduction Why do some people collapse under life st resses while others seem unscathed by traumatic circumstances such as severe illness, the death of loved ones, and extreme poverty, or even by major catastrophe s such as natural disasters and war? Surprisingly large numbers of people mature into normal, successful adults despite stressful, disadvantaged, or ev en abusive childhoods. Yet, other people are so emotionally vulnerable that se emingly minor losses and rebuffs can be devastating. (Basic Behavioral Science Task Force of the National Advisory Mental Health Council, 1996, p. 22) Life is a complex web of pros and c ons, positive and negative experiences, protective and risk factors, and gains and losses. Crimi nological theory addresses why people with similar or identical life experiences vary in their willingness to adhere to the proscribed laws of society. In partic ular, some criminologists (Agnew, 1992, 2001; Bernard, 1987; Cloward, 1959; Cloward & Ohlin, 1960; Cohen, 1955, 1965; Durkheim, 1897/1951; Merton, 1938, 1968; Messner, 1985, 1988; Messner & Rosenfeld, 1994), have postulated theoretical explanations hypot hesizing how life stresses interfere with goal attainment, and why certain individuals or groups cope with this interference through legitimate strategies, while others use illegitimate coping strategies to circumvent or relieve stress. The most dominant of these theories are anomie theory, classic strain

PAGE 13

2 theory, institutional anomie theory, and ge neral strain theory. While the specific assumptions and propositions of these theories differ, each suggests that stress or strain plays an important role in the etiology of crime/deviance. Anomie theory is a macro-level or structural theory of crime. Anomie is a term first used by Durkheim (1893/1964, 1897/1951) to describe the inability of society to maintain order or regulation over the desires and aspirations of its members. Durkheim suggested that under conditions of increased societal change (e.g., industrial growth, financial crisis) a state of “normlessness” occu rred within a society th at left it temporarily unable to enforce laws and rules among its peop le. With the social order weakened, the people may resort to unconventional means of achieving their desires, and crime rates increase. Merton relied heavily upon Durkheim’s work to articulate a more culturally driven and formal version of anomie theory (Merton, 1938). Merton claimed that society must maintain a balance between culture (soc ially approved goals) and social structure (socially approved means). If every person in the culture is expected to strive for the same goals, but not provided equal structural means (i.e., status), then anomie is more likely to result. When an imbalance exists between culture and social structure and the overall goals of society, there is an increas ed likelihood that anomie will result at the societal level and strain will result at the group and individual level (Kornhauser, 1978, p. 143). Strain is defined in Mertonian terms as pressure or frustr ation (i.e., stress) on cultural groups to achieve socially defined economic success. Strain is considered a “mode of adaptation” to anomie. It is measured as the imbalance between economic

PAGE 14

3 aspirations and expectations. Merton’s anomie theory became the basis of classic strain theory and other versions of strain theory. Classic strain theory relied heavily on the contributions made by Cohen (1955, 1965), Cloward (1959), and Cloward and Ohlin (1960). Like Merton, Cohen, Cloward, and Ohlin acknowledged the existence of macr o-level anomie, however, they focused on lower-class juvenile subcultu res in particular. Cohen suggested that subcultural delinquency among lower-class boys was caused by strain induced by blocked goals of status and social acceptance, rather than goals of economic success. He believed working-class boys strive for mi ddle-class status and respect, and that “status frustration” or strain is experienced when this goal is blocked. J uvenile subcultures experiencing high levels of status strain are more likely to engage in higher rates of delinquency. Cloward and Ohlin (Cloward, 1959, Cloward & Ohlin, 1960) believed that Cohen was not correct in his assumption that the workingclass strives for status achievement rather than economic achievement. Similar to Me rton (1938), they hypothesized that workingclass boys, specifically delinque nt gangs, were driven by economic goals. Juvenile subcultures that experienced blocked opportunities for economic success were more likely to engage in higher crime rates. Ho wever, Cloward and Ohlin (1960) also suggest that crime rates depended upon access to i llegitimate opportunities, denied access to legitimate opportunities was not sufficient to produce delinquency. Due to the focus of these theories on juvenile gang behavior, crimi nologists have misinterpr eted this to mean that classic strain theory is applicable to the explanation of individual difference in crime (for detail see Bu rton & Cullen, 1992).

PAGE 15

4 Classic strain theories have been critic ized for a variety of reasons. However, three major criticisms have emerged. First, classic strain theory has received little empirical support (e.g., Akers & Cochran, 1985; Burton, 1991; Burton, Cullen, Evans, & Dunaway, 1994; Elliott, Huizinga, & Aget on, 1985; Hirschi, 1969; Johnson, 1979; Liska, 1971; Quicker, 1974; Voss, 1966; but see Fa rnworth & Lieber, 1989). According to strain theory, crime should be highest when aspirations for success were high and expectations were low. However, most studie s of strain theory have indicated that crime is highest when both aspirations and exp ectations are low, and lowest when both aspirations and expectations are high (s ee Hirschi, 1969; Kornhauser, 1978). Second, strain theory assumes that crime will be c oncentrated in the lower-class because in the lower-class goals are overemphasi zed at the expense of means. Yet, studies have shown that the middle-class experiences high crime, and that class is weakly related to crime (e.g., Hindelang, Hirschi, & Weiss, 1981; Kr ohn, Akers, Radosevich, & Lanza-Kaduce, 1980; Thornberry & Farnworth, 1982; but see Elliott & Huizinga, 1983). Third, classic strain theory has been criticized because it does not provide an explanation for desistence and periods of criminal inactivity among youths (Hirschi, 1969). Based on these criticisms, social scientists have proposed th eoretical revisions to classic strain theory (see Agnew, 1992; Burton & Cullen, 1992; Farnworth & Leiber, 1989; Jensen, 1995; Messner & Rosenfeld, 1994). In general, revisions of classic strain th eory can be characterized as belonging to one of two types: structural or individual. Struct ural revisions of cl assic strain theory remain true to the macro-level hypothes is of classic stra in theory that anomie or structural strain (i.e., blocked opportunities to achieve m onetary success and/or middle-class status)

PAGE 16

5 is a cause of the rate of crime (e.g., Be rnard, 1987; Messner, 1985, 1988). Messner and Rosenfeld’s (1994) institutional anomie theory is among the most notable macroclassic strain theory revisions. Institutional anomie theory suggests that the American economy dominates all other social inst itutions, such as the educationa l system, the family, and the political system (Messner & Rosenfeld, 1994) In a balanced society, non-economic social institutions serve to insulate society’s members from crime. Under the ideology of the American Dream, however, disproportio nately high crime ra tes result from the overemphasis placed on the economic institution. This structural revision of strain theory has received a respectable amount of em pirical support (Chamlin & Cochran, 1995; Messner & Rosenfeld, 1997; Piquero & Piquero, 1998; Pratt & Godsey, 2003; Savolainen, 2000). Individual level revisions of classic strain theory have shifted the focus of the theory from a structural or macro-level persp ective to a micro-level, social-psychological perspective (Agnew, 1992; Burton & Cullen, 1992) in an effort to better conceptualize the theory. Robert Agnew’s Ge neral Strain Theory is the most notable of the micro-level revisions of classic strain theory. General strain theory has recei ved much consideration in recent years and acquired a respectabl e amount of empirical support (e.g., Agnew, 2002; Agnew & Brezina, 1997; Agnew & White 1992; Baron & Hartnagel, 1997, 2002; Benda & Corwyn, 2002; Brezina, 1999; Broidy, 2 001; Eitle, 2002; Eitle & Turner, 2002, 2003; Hoffmann, 2002; Hoffmann & Cerbone, 1999; Hoffmann & Miller, 1998; Hoffmann & Su, 1997; Maxwell, 2001; Maze rolle, 1998; Mazerolle & Maahs, 2000; Mazerolle & Piquero, 1997, 1998; Paternoster & Mazerolle, 1994; Piquero & Sealock, 2000, 2004; Robbers, 2004).

PAGE 17

6 General strain theory offers a modified con ceptualization of strain, such that strain is now defined as “ negative relationships with others : relationships in which the individual is not treated as he or she wants to be treat ed.” (Agnew, 1999, p. 48). This new conceptualization broadens the definiti on of strain by incorporating more complex dynamics related to positive and negative stim uli of stress, thus allowing for a more diverse measurement of how strain can occur. Specifically, according to GST, strain can be conceptualized as being co mprised of three forms of strain: (1) failure to achieve positively valued goals, (2) removal of positively valued stimuli, and (3) presentation of negative stimuli (Agnew, 1992). GST hypothesizes that when individuals fail to achieve positively valued goals (i.e., educational, income, and status derived immediate and longterm goals) they experience frustration or pre ssure, which may be more likely to lead to crime. In addition, individuals may experience strain when positive stimuli (e.g., relationships with loved ones) are removed from their lives Removal of positive stimuli can increase frustration, which, in turn, may in crease the changes that crime will result. Individuals may also experien ce strain when negative or noxious stimuli (e.g., negative relationships with parents and teachers such as abuse or neglect) ar e introduced in their lives. This negative stimulus cr eates a pressure or frustration to allevi ate or remove the negative stimuli, which may increase the likelihood that crime will result. General strain theory (Agnew, 1992, 2001) posits that an individual will experience at least one negative emotion, referred to as negative affect per experience of strain. Negative affect refers to negative emotional states (e.g., depression, anxiety, and anger) that emerge due to the frustration caused by strain. However, not everyone experiencing strain or negative affect will co mmit crimes. Whether or not negative affect

PAGE 18

7 leads to an illegitimate response depends on th e development and presence of individual coping strategies (i.e., cognitive, emotiona l, and behavioral adaptations) and other conditioning factors (e.g., intelligence, interper sonal skills, social support systems) that are present and available for acce ss by the individual. Since the introduction of gene ral strain theory, Agnew a nd others have attempted elaborate on the comprehensiveness of the th eory, making it more and more general in its application to the etiology of crime. Ofte n these theoretical expansions of GST have been guided by the findings of previous studi es. These expansions have provided more specification of criminal motivations (A gnew, 1992), criminogenic types of strain (Agnew, 2001), gender differences (Broidy & A gnew, 1997), structural effects that may condition the strain-crime relationship (A gnew, 1999), developmental or life-course differences in strain (Agnew, 1997), and biol ogical explanations of the strain-crime relationship (Walsh, 2000). Until recently, the GST literature has ignored what Agnew and associates (Agnew, Brezina, Wright, & Cullen, 2002) ar gue may be one of the most important conditioning effects of the strain-crime relatio nship, namely the personality traits of the individual experiencing stra in. In his foundation for GST, Agnew (1992, p. 65) alluded to the role that personality may play in GST in his discussion of conditioning factors influencing the strain-crime relationship, t hough no specific mention of personality traits or psychopathic features was made. Agne w suggested “temperament,” a less stable precursor to personality traits (Goldsmit h, 1996; Pedlow, Sanson, Prior, & Oberklaid, 1993; Rothbart & Bates, 1998), may serve as moderating factors for the straindelinquency relationship. Several years late r, Agnew stated that “[t]he subjective

PAGE 19

8 evaluation of an objective strain is a function of a range of factors, including individual traits (e.g., irritability )…” (Agnew, 2001, p. 321). Personality traits are relatively stable characteristics that describe one’s perception and behavior toward the envir onment (Caspi, Moffitt, Silva, StouthamerLoeber, Krueger, & Schmutte, 1994). Ther e is impressive evidence that suggests personality traits may be stable and enduri ng characteristics, aff ected by biological and early socialization processes (Bock & Goode, 1996; Carey & Goldman, 1997; Eley, 1998; Gottesman & Goldsmith, 1994; Lykken, 1995; Moffitt, 1987; Plomin & Nesselrode, 1990; Rutter, 1996; see also, Wals h, 2000). The literature has consistently revealed a significant association between pe rsonality traits that are non-conforming or maladaptive and aggression and antisocial behavior among adult and juvenile samples (e.g., Binder, 1988; Blackburn & Coid, 1998; Ca spi et al., 1997; Caspi et al., 1994; Cloninger, 1987; Eysenck & Eysenck, 1985; Eysenck & Gudjonsson, 1989; Farrington, 1986, 1992; Hare & Jutai, 1983; Harris, Rice & Cormier, 1991; Hart, Kropp & Hare, 1988; Hemphill, Hare & Wong, 1998; Kosson, Smith & Newman, 1990; Luengo, Otero, Carrillo-de-la-Pea, & Mir n, 1994; Mak, Heaven, & Rummer y, 2003; Miller & Lynam, 2001; Miller, Lynam, Widiger & Leukef eld, 2001; Raine, 1993; Robins, 1966; Rutherford, Alterman, Cacciola & McKay, 1997; Rutter & Giller, 1983; Salekin, Rogers & Sewell, 1996; Smith & Newman, 1990; Tenn enbaum, 1977; Tremblay, Pihl, Vitaro, & Dobkin, 1994; Wilson, Rojas, Haapanen, Duxbur y, & Steiner, 2001; Zuckerman, 1989). According to Miller and Lynam (2001), th e following overview can be made about personality and crime:

PAGE 20

9 Individuals who commit crimes tend to be hostile, self-centered, spiteful, jealous, and indifferent to others.…They tend to lack ambition, motivation, and perseverance, have difficulty controlli ng their impulses, and hold nontraditional and unconventional values and beliefs. (p. 780) Given the relationship between certain personality traits and aggression and crime, Agnew et al. (2002) have suggested that personality traits may be important moderators of the effect of st rain on crime. Personality trai ts may affect how individuals emotionally respond to strain and develop c oping strategies to st rain. Individuals possessing maladaptive personality traits are hy pothesized to interpret strain as aversive and are more likely to experience negative a ffect in the form of anger (Agnew et al., 2002, pp. 45-47). Such individuals are also more likely to perceive aggressive solutions to strain as better coping mechanisms for their situations (pp. 45-47). Based on these assumptions, Agnew et al. (2002) examined how strain is moderated by individual personality traits characteris tics. The authors determined that certain features of personality (negative emotionality and low cons traint) moderate the effect of strain on delinquency. By highlighting the role that personality traits may play in GST, Agnew and associates have provided an opportunity for a more complete explanation of how strain motivates deviance. Through an empi rical examination of conditional factors, Agnew is attempting to provide a more generalizable theory of crime. Agnew’s recent article is noteworthy not simply because of its more generalized application, but because this study has allo wed for the incorporation of a whole new perspective in strain theo ry. Accordingly, this new framework emphasizes the psychological aspects of the theory. At presen t, there is an abundance of mainstream and

PAGE 21

10 academic interest in the psychology of crime, especially maladaptive personality traits like psychopathy (Campbell, Porter, & Sa ntor, 2004; Catchpole & Gretton, 2003; Corrado, Vincent, Hart, & Cohen, 2004; Falken bach, Poythress, & Heide, 2003; Gretton, McBride, Hare, O’Shaughnessy, & Kumka, 2001; Kosson, Cyterski, Steuerwald, Neumann, & Walker-Matthews, 2002; Lee, Vincent, Hart, & Corrado, 2003; Lynam, 1997; Lynam et al., in press; Murrie & Cornell, 2002; Murrie, Cornell, Kaplan, McConville, & Levy-Elkon, 2004; O’Neill, Li dz, & Heilbrun, 2003; Pardini, Lochman, & Frick, 2003; Spain, Douglas, Poythress, & Epstein, 2004; Stafford & Cornell, 2003; Vitacco, Rogers, & Neumann, 2003; Ridenour 2001; Salekin, Leistico, Neumann, DiCicco, & Duros, 2004; Vitacc o et al., 2003). Therefore, Agnew and colleagues’ (2002) work presents a timely, relevant, and substant ive contribution to the field. In an attempt to build upon their work and make n addi tional contribution to the discipline of criminology, the present study provides a replic ation and extension of the Agnew et al. (2002) test of general strain and personality traits. Toward this end, Chapter 2 presents the theoretical foundation for general strain theory (Agnew, 1992, 2001). Key theoretical c oncepts are defined, and an overview of the empirical support for the framework is pr esented. Many of the key concepts for GST overlap with concepts describe d in social control (Hirschi 1969; Gottfredson & Hirschi, 1990) and social learning theories (Ake rs, 1973, 1977, 1985; see also, Burgess & Akers, 1966; Sutherland, 1947). Therefore, Agnew argue s that tests of GST should control for measures of social control and differentia l association. The importance of examining rival theoretical measures, social control and social learning theory, when conducting a

PAGE 22

11 full test of GST is also discussed. In relati on to this issue, the chapter concludes with a discussion of tautology and fa lsifiability in GST. The purpose of the present study is to exam ine the role that personality plays in GST. Therefore, before existing empirical lit erature testing the eff ects of personality on strain (i.e., the 2002 Agnew et al. article) and the hypotheses of this study can be presented, if is critical to examine what th e literature says about personality and its relationship to antisocial beha vior. Chapter 3 presents an overview of personality traits and psychopathy personality traits or psychopa thy. The chapter includes a discussion of the temporal consistency and stability of personality. Empirical evidence of an association between personality traits a nd psychopathic features and delinquency, substance use, and strain is presented. The present study examined GST and personality among a justice-referred sample of adolescen ts; therefore, recent studies that have extended the concept of psychopathy downward from adults to children and adolescents are also presented. Moreover, tautological issues in the measurement of personality and psychopathy compared to antisocia l behavior are considered. Chapter 4 presents a brief discussion of A gnew et al.’s (2002) test of GST and the moderating effects of personal ity traits, specifically negative emotionality and low constraint. The proposed study, a replication and extension of the Agnew et al. article, is presented. Models presenting the structur al equation models analyses that were conducted in this study are i llustrated and explained. Chapter 5 describes the sample used in this study. Details regarding the operationalization of the strain, social cont rol, differential association, personality, delinquency, and drug problems variables are prov ided. Chapter 6 explains the analytical

PAGE 23

12 strategy employed in this study. Finally, Chap ter 7 contains a disc ussion of the findings, limitations, and implications for policy and future research.

PAGE 24

13 Chapter 2 General Strain Theory Strain theory (see Merton, 1938, 1968; Cohen, 1955; Cloward & Ohlin, 1959, 1961) is among one of the more venerable soci ological and criminologi cal theories (Cole, 1975). Classic strain theory was originally derived from Emile Durkheim’s anomie theory (1897/1951), although Robe rt Merton (1938) is most often credited with the modern-day conceptualization of anomie theo ry. Anomie theory attempts to explain societal variations in crime rates, and as such describes a state of “macrosocial disorganization” (Kornhauser, 1978) or “normlessness” (Durkheim, 1897/1951) that leads to higher levels of crime. Since the early 1900s, anomie theory has become narrower in scale, explaining w hy certain groups of individuals within societies, rather than entire societies, have higher criminal tendencies than others (Cloward & Ohlin, 1960; Cohen, 1955; Merton, 1968). These offs hoots of anomie theory form the conceptual basis of classic strain theory. Anomie and Classic Strain Theory Anomie is a concept first used by the 19th century French sociologist, Emile Durkheim, to describe the inability of soci ety to maintain order or regulation over the desires and aspirations of its memb ers (Durkheim, 1893/1964, 1897/1951). Durkheim suggested that a state of “nor mlessness” occurred within a society when dramatic changes or disruptions left the soci ety temporarily unable to en force common laws and rules

PAGE 25

14 among its people. Examples of these disrupt ive changes could include such events as financial crisis and rapid industrial grow th (Passas, 1995). Under these conditions consequences of anomie may be observed, such as increases in competitiveness, greed, status aspirations, and pleasur e-seeking (Passas, 1995). With the social order weakened, the people may resort to unconventional mean s of achieving their anomic desires. Therefore, deregulated societies may experi ence temporarily higher levels of crime until a sense of order can be re-established. Merton relied heavily upon the theoretical framework established by Durkheim to propose a more culturally driven explanati on of anomie (Merton, 1938). In particular, Merton’s Social Structure and Anomie th eory (1938) provided an explanation of deviance in American society. Merton claime d that society must maintain a balance between culture and social structure. Cultu re refers to the values that characterize appropriate goals (i.e., aspira tions) and means. An integr ated culture equally stresses both goals and means. However, “malinte grated” cultures overemphasize goals and/or means. Cultures that overemphasize goals at the expense of means will be more likely to experience anomie. Cultures that overemphasi ze means at the expense of goals will be more likely to be ritualistic and rigid in nature Social structure refers to the presence or absence of social class stratifi cation. Societies that possess stratified social structures (i.e., divisions in social status or class) have structural in equalities in the distribution of means. Among egalitarian socie ties, especially, every member is expected to aspire to the same goals, regardless of social class. If every person in the culture is expected to strive for the same goals, but not provided equal structural means (i.e., status), then anomie is more likely to result. When a co mbined imbalance or “malintegration” exists

PAGE 26

15 between culture and social structure, there is an overall increased likelihood that anomie will result at the societal level and strain will result at the group and individual level (Kornhauser, 1978, p. 143). Strain is define d in Mertonian terms as pressure or frustration on cultural groups to achieve so cially defined economic success. It is measured as the disjunction between econom ic aspirations and expectations. Within American society, the pur suit of the “American Dream” causes “malintegration” of culture and social struct ure, which leads to unusually high crime rates (Merton, 1938, 1957). The vision of the Amer ican Dream substantially overstresses goals of economic and status su ccess at the expense of mean s. Americans are taught by family, friends, school, the media, and others to attain financial succ ess and notoriety at almost any cost. Little is said of the appropriate means to achieve these goals. Moreover, disadvantaged minority groups are ex pected to internalize the same goals as more affluent groups, despite little if a ny legitimate opportunities for achieving these goals. The pursuit of the American Dream lead s to such an imbalance within culture and social structure that anomie is produced. This anomic condition creates strain among groups and individuals to achieve the American Dream regardless of so cial prohibitions. As a result, many strained Americans pursue success through illegitimate means, which may be one possible explanation of why Amer ica exhibits much higher crime rates than other countries (Akers, 1997). Laying the groundwork for classic strain theory, Merton (1938) described five cultural adaptations to strain, four of whic h are deviant. These “modes of adaptation” differ depending upon the emphasis that is placed on the balance of goals to means. First, most cultural group members may respond to strain through conformity Conformists

PAGE 27

16 continue to strive for economic success thr ough legitimate channels regardless of anomic and strainful situations they experience. Second, group members may respond to strain through innovation Groups reacting to strain through innovation have a high commitment to societal goal s, but a low commitment to conventional or legitimate means. They will utilize illeg itimate means to attain their goals when necessary. Third, some members may respond to strain through ritualism These members demonstrate lower commitment to goals, but a zealous co mmitment to legitimate means. Fourth, another type of reaction to strain is rebellion Rebellious members are not committed to either the goals or legitimate means of society, but rather, ve rsions of their own goals and means. Such groups will be highly committe d to their own versions of goals and means (e.g., revolution, political uprising) in place of those of conventional society. Finally, retreatism is another form of strain adaptation. Like rebellious members, retreatists will reject the goals and means of society. Howeve r, they will not substitute their own goals and means for those of society; instead, they surrender and dropout of society altogether. Although Merton presented the five “modes of adaptation” as structural responses to strain, Menard (1995) has suggested that thes e cultural adaptations to strain can be attributed to the individual level. The aforementioned suggests that bloc ked cultural goals and easy access to illegitimate means of success may lead to social processes conducive to anomie and deviant behavior, thus establis hing the foundation for classic stra in theory (Passas, 1995). The transition from Merton’s anomie theory to classic strain theory relied heavily on the contributions made by Cohen (1955, 1965), Cloward (1959), and Cloward and Ohlin (1960). Like Merton, Cohen, Cloward, and Ohli n acknowledged the existence of societal

PAGE 28

17 anomie; however, their focus on juvenile s ubcultures expanded the focus of anomieinducing factors beyond economic st rain (Passas, 1995, p. 101). Albert Cohen (1955, 1965) applied of strain theory to the study of juvenile crime variations. In his struggl e to understand juvenile crim e, Cohen applied Merton’s concepts of structural (i.e., cu ltural) sources of strain to th e lower-class, urban adolescent male subculture. Cohen suggested that subcultural delinquency among lower-class boys was caused by strain induced by blocked goals of status and social acceptance, rather than goals of economic success. That is, working-class boys strive for middle-class status and respect. When this goal is blocke d, they experience “status frustration” or strain, which may lead to assi milation of a delinquent subculture. Acceptance into the delinquent subculture is achieved through st atus mobility within delinquent gangs. Cohen’s strain theory of delinquent gangs ha s been criticized as relying so heavily on delinquent subculture that it “hovers on the brink of adopting a cultural deviance explanation of working-class delinquency,” ra ther than a strain e xplanation (Kornhauser, 1978, p. 154). Cloward and Ohlin (Cloward, 1959, Clowar d & Ohlin, 1960) presented a strain theory of delinquent subcultures that is more in line with Merton’s original theory and supportive of Cohen’s strain theory. Cl oward and Ohlin believed that Cohen was incorrect in his assumption that the workingclass strives for status achievement rather than economic achievement. They believed that Merton was correct in his assumption that Americans desire money and economic success; however, he was wrong in assuming that it was the imperfect socialization of the working-class that led to higher rates of crime among this culture (Kornhauser, 1978, p. 156). Cloward and Ohlin suggested that

PAGE 29

18 working-class boys, specifically delinquent gangs, were driven by economic goals and just as capable of conformity as middle-cla ss or higher-class Americans. The difference was that working-class boys were simply conforming to the norms and beliefs of a different subculture. Moreover, denied access to legitimate opportunities was not sufficient to produce delinquency; delinquent gangs also had to ha ve access to illegitim ate opportunities. Borrowing from Shaw and McKay’s (1942) social disorganization theory and Sutherland’s (1947) differential association theory, Cloward and Oh lin suggested that cultural transmission of delinque nt values and differential opportunities to illegitimate means resulted in deviant adaptation when economic goals were blocked (Akers, 1997). Delinquent gangs varied in their level of delinquent involvement or specialization. Differential opportunities to illegitimate means explained why criminal conflict and retreatist gangs specialized in theft, fighting, an d alcohol/drug use, respectively (Cloward & Ohlin, 1960). Criticism of Classic Strain Theory As mentioned above, Merton proposed a macro-level theory of anomie, which included assumptions about how societal conditions create macro-level strain toward anomie (Merton, 1968). However, scholars (Agnew, 1992; Hirschi, 1969; Kornhauser, 1978) began to reinterpret Merton’s notion of strain as a micro-level explanation of strain. In part, this misinterpretation of st rain is a consequence of Merton’s own writings, in which he makes reference to the effects of strain upon individua ls (Burton & Cullen, 1992). Burton and Cullen said it best when they stated the following:

PAGE 30

19 [I]f Merton wanders into the realm of the individual, ultimately he retreats from this level of analysis and reminds us that anomie is a societal condition and that his theoretical purpose is fundamentally sociological: to explain rates of deviance/crime across the social structure, not to explain whic h individuals feel the pressure to engage in such wayward activities. (p. 5) The works of Cohen (1955, 1965) and Cloward and Ohlin (Cloward, 1959, Cloward & Ohlin, 1960) may have also contributed to the attr ibution of strain to the individual level. They focused so heavily upon understandi ng delinquency among working-class boys (i.e., gangs) that scholars may have misint erpreted this to mean that individual differences in delinquency within gangs should be examined (Burton & Cullen, 1992, p. 6). In actuality, Cohen and Cloward and Ohlin were suggesting that gang group differences in delinquency within urban areas should be examined (p. 6). Further, the conceptualization of strain theory in crit icisms made by Hirschi (1969) and Kornhauser (1978) may have contributed in large part to the social psychological and microinterpretations of anomie theory (Burton & Cullen, 1992, p. 6-7). Where the responsibility for the microlevel interpretation of anomie theory, which has become known as classic strain theory, lies is irrelevant in this particular study. What does matter is that tests of this classic strain theory received marginal empirical support. Classic strain theory has tradit ionally been tested by examining (a) the disjunction between succe ss (e.g., occupation, educational, economic measures) aspirations and expectations and (b) blocke d opportunity (Burton & Cullen, 1992). The empirical literature examining the disjunction between aspirations a nd expectations have suggested that classic strain theory is empirically w eak (e.g., Akers & Cochran, 1985;

PAGE 31

20 Burton, 1991; Burton, Cullen, Evans, & Dunaway, 1994; Elliott, Huizinga, & Ageton, 1985; Hirschi, 1969; Johnson, 1979; Liska, 197 1; Quicker, 1974; Voss, 1966; but see Farnworth & Lieber, 1989). The empirical li terature testing clas sic strain theory by measuring perceptions of blocked opportuni ties has provided mixed support for the theory (e.g., Agnew, 1984; Burton, 1991; Cernkovich & Giodano, 1979; Segrave & Halsted, 1983). The weak empirical support fo r classic strain theory has resulted in significant criticism from within the sociolog ical discipline (Akers 1996; Hirschi, 1969; Kornhauser, 1978). While classic strain theories have been criticized for a variety of reasons, three major criticisms have emerged. First, as me ntioned, classic strain theory has received little empirical support. According to stra in theory, crime should be highest when aspirations for success were high and expectatio ns were low. However, most studies of strain theory have indicated that crime is highest when both aspira tions and expectations are low, and lowest when both aspirations a nd expectations are high (see Hirschi, 1969; Kornhauser, 1978). This has been interprete d as supporting a social control perspective rather than a strain theory perspective (see Hirschi, 1969; Kornhauser, 1978). Second, strain theory assumes that crime will be concentrated in the lower-class because here goals are overemphasized at th e expense of means. Yet, studies have shown that the middle-class experienced high crime, and that class is weakly related to crime (e.g., Hindelang, Hirschi, & Weiss, 1981; Krohn, Akers, Radosevich, & Lanza-Kaduce, 1980; Thornberry & Farnworth, 1982; but see Elliott & Huizinga, 1983) Finally, classic strain theory has been criticized because it does not provide an explanation for desistence and periods of criminal inactivity among youths (Hirschi, 1969). Based on these criticisms,

PAGE 32

21 social scientists have proposed theoretical revisions to strain theory (see Agnew, 1992; Burton & Cullen, 1992; Farnworth & Leiber, 1989; Jensen, 1995; Messner & Rosenfeld, 1994). Revisions of classic strain theory conte nd that the key to improving the empirical adequacy of the theory lies in the clar ification of its c onceptualization and operationalization—and the specif ication of it. Advocates fo r these revisions argue that previous tests of strain theo ry have inadequately measured the concept of strain (e.g., Agnew, 1992; Berton, 1987; Burton & Culle n, 19992; Cullen, 1984; Farnworth & Leiber, 1989; Messner, 1988). In general, revisions of classic strain theory can be characterized as belonging to one of two types: structural or individual. Structural revisions of classi c strain theory are actually attempts to return to the original theoretical premise proposed by Merton in his anomie theory. These revisions remain true to the macro-level hypothes is of classic stra in theory that anomie or structural strain (i.e., blocked opportunities to achieve m onetary success and/or middle-class status) is a cause of the rate of crime (e.g., Bern ard, 1987; Messner, 1985, 1988). In particular, Messner and Rosenfeld (1994) have revised anomie/strain theory into a macro-level theory of anomie called instit utional anomie theory. Institu tional anomie theory purports that the American social “i nstitution”, namely the economy, dominates all other social institutions, such as the educational system, the family, and the political system (Messner & Rosenfeld, 1994). In a balanced society, non-economic social institutions serve to insulate society’s members from crime. Under the ideology of the American Dream, however, disproportionately high crime rates result from the overemphasis placed on the economic institution. This structural revision of strain theory has received a respectable

PAGE 33

22 amount of empirical support (Chamlin & Cochran, 1995; Messner & Rosenfeld, 1997; Piquero & Piquero, 1998; Pratt & Godsey, 2003; Savolainen, 2000). Recent revisions of the strain tradition have shifted the focus of the theory from a structural or macro-level perspective to a micro-level, social-psy chological perspective (Agnew, 1992; Burton & Cullen, 1992) in an effo rt to better conceptualize the theory. One such revision of strain theo ry in particular has received much consideration in recent years: Robert Agnew’s Genera l Strain Theory (1992). General Strain Theory Within a traditional micro-social context of strain theory, strain is defined as the frustration of desires, needs, or wants. Based on this definition, strain is operationalized as the difference between what is desired (i .e., aspirations) and the anticipated outcome (i.e., expectations) and/or as the differe nce between the anticipated outcome or expectations and the actual outcome obt ained. Delinquency is motivated by the anticipated gratificatio n of frustrated desires. Genera l strain theory elaborates on the conceptualization and operationaliza tion of classic strain theory. According to general strain theory, strain is defined as “ negative relationships with others : relationships in which the individual is not treated as he or she wants to be treated” (Agnew, 1992, p. 48). General strain theory posits that an individual will experience at least one negative emotion, referred to as negative affect per experience of strain. Negative affect may span a broa d spectrum of negative emotional states, including such expressions as depression, anxi ety, and anger. More specifically, Agnew argues that anger is perhaps th e most important form of negative affect and serves as a key motivator to strain-induced deviance. Acco rding to Agnew, anger is one of the most

PAGE 34

23 potent reactive emotions due to its tendenc y to produce a desire for retribution. He argues that individuals are “… pressured into delinquency by the negative affective states—most notably anger and related emo tions—that often result from negative relationships …” (p. 49). However, Agnew recogni zes that not everyone experiencing strain or negative affect will commit crimes. Whether or not negative affect leads to an illegitimate response depends on the devel opment and presence of individual coping strategies and other conditioning factor s that are conducive to such action. Types of Strain General strain theory delin eates three major types of strain that may lead to delinquent or criminal behavior1 (Agnew, 1992): (1) failure to achieve positively valued goals, (2) removal of positively valued stimuli, and (3) presentation of negative stimuli. Strain as the inability to achieve positively valued goals is subdivided into three categories. The first sub-category refers to strain as a disjuncti on between aspirations (i.e., ideal goals) and expecta tions (i.e., anticipated or ac tual goals) (Agnew, 1992, p. 52). That is, strain is caused by incongruence betw een one’s ideal goals and one’s anticipation of actual goals. Within this sub-category these ideal goals are typically culturally derived. For example, a youth who comes from a family with very limited financial means (i.e., lower class socioeconomic status) may experience strain when he aspires to receive an expensive car from his parents for his sixteenth birthday, and then does not receive the car. Individuals may engage in illicit acts to overcome an experienced gap between aspirations and expectations. 1 Although general strain theory is postulated to be “general” in its application and explanation of deviant behavior, much of the research on general strain theory pertains to adolescents. For the purposes of this paper, both criminal and delinquent beha vior will henceforth be referred to as delinquency in this context.

PAGE 35

24 This sub-category encompasses strain as described and measured by the earlier micro-social version of classic strain th eory (see Merton, 1968; Cohen, 1955; Cloward & Ohlin, 1959, 1961). Criticism and a lack of strong empirical support for such a delineation of strain (see Agnew, 1991; Bern ard, 1984; Burton et al., 1994; Elliott, Huizinga, & Ageton, 1985; Farnworth & Leib er, 1989; Kornhauser, 1978; Liska, 1987; also, for an explanation as to why this sub-ca tegory of strain is less likely to affect crime/delinquency see Agnew, 2001), however, ha s led to a revision of this sub-category that emphasizes more immediate aspirations and expectations. Agnew suggests that certain youth subcultures emphasize more im mediate goals (e.g., getting good grades, popularity) versus long-term goals (e.g., careers, college). He argues that consideration of more immediate goals is particularly impor tant when examining juvenile behavior and delinquency (Agnew, 1992, p. 51). It should be noted that even with such consideration of more immediate goals, strain as the disj unction between aspiratio ns and expectations has received weak empirical support (see Agnew, 2001). The second sub-category of strain deve loping from an inability to achieve positively valued goals results from the disj unction between expectations (rather than ideal goals) and actual achievements (Agnew, 1992, p. 52). In other words, strain is caused by a gap between one’s expected goals and one’s actual achievements. The previous sub-category of strain, aspirations versus expectations is based on ideal circumstances, representing rather utopian id eals. Agnew suggests, however, that more realistically grounded goals should be more st rain-inducing than idealistic aspirations. Expectations are formulated from a person’ s past experiences and comparisons with similar others (i.e., referentia l others). They provide a mo re realistic evaluation of an

PAGE 36

25 individual’s capabilities. In this situation, for example, an athletically built high school junior, who has previously pa rticipated in other sports, may reasonably expect to be selected for the varsity football team, but then experiences strain when he is not selected. In an effort to overcome the frustration th at results from an experienced gap between expectations and achievements, indi viduals may engage in illicit acts. The third sub-category of strain originating from an inability to achieve positively valued goals defines strain as the disjuncti on between just or fair outcomes and actual outcomes (Agnew, 1992, pp. 53-55). According to this measure of strain, individuals expect a certain degree of equality or distribut ive justice in the allocation of resources. This form of strain is perhaps best conceive d of as a scale weighing the amount of efforts extended compared to the re wards reaped. When the amount of effort expended is equivalent in magnitude to that of the outco me, the relationship is considered “just” or “fair”. On the other hand, if the size of the e ffort is greater than th at of the outcome, the relationship is considered “unjust” or “unfair”.2 For instance, two students study together for the same exam in the exact same manner. One student receives an “A”, while the other receives a “D”. The student receiving the lower grade may feel that the outcome was unjust if she also believes that in all other respects she and th e other student were similar (i.e., no mitigating circumstances, such as intelligence, that affected the outcome). Individuals in inequitable relationships may e ngage in delinquency to shift the balance of equity in their favor. It is important to note that this last sub-category of stra in is considered particularly significant for GST, and is hypothe sized to be one of the more criminogenic 2 It is rare that an individual will experience strain as a result of a relationship in which the amount of effort put forth is less than the outcome received.

PAGE 37

26 forms of strain (see Agnew, 2001, p. 327). Inde ed, the concept of in justice is not limited to merely this one sub-category of strain, but is applicable to all types of strain. Studies have shown a strong association between pe rceived injustice and anger (Agnew, 1992; Averill, 1982, 1993; Berkowitz, 1993; Tedesc hi & Felson, 1994; Tedeschi & Nesler, 1993; Tyler, Boeckmann, Smith, & Huo, 1997), wh ich has been demonstrated to be an antecedent of delinquent behavior (Aselt ine, Gore, & Gordon, 2000; Berkowitz, 1993; Brezina, 1998; Mazerolle, Burton, Cullen, Evan s, & Payne, 2000; Mazerolle & Piquero, 1998; Piquero & Sealock, 2000; Tedeschi & Felson, 1994). The second major type of strain is ca used by the removal of positively valued stimuli (Agnew, 1992, pp. 57-58). Again, Agnew formulates this concept based on the stress literature, which indicates that when previously administered positive stimuli are reduce or withheld, aggression follows (Bandura, 1973). Examples of this type of strain include inventories of stressful life events containing items su ch as the death of a loved one, the loss of a close friend or significan t other, and divorce of one’s parents. Delinquency may result when an individual attempts to regain all or portions of a lost or blocked positive stimulus, seek revenge on those causing the loss of a positive stimulus, and/or cope with a lost positive stimulus by using illicit substances (Agnew, 1992, pp. 57-58). The third major type of strain is cause d by a confrontation with negative stimuli (Agnew, 1992, pp. 58-59). Noxious or negative s timuli are powerful situations that the individual, particularly adolescents, cannot easily avoid. Examples of such negative stimuli include abuse (physical, sexual, a nd/or emotional) from a parent, negative relations with teachers or other adults, and physical or other threats from peers (i.e.,

PAGE 38

27 bullying, teasing). Agnew relies on the stress literature that indicates aggression and other negative consequences may follow the pr esentation of negative st imuli. As a result, Agnew hypothesizes negative stimuli may lead to delinquency as an adolescent attempts to avoid (i.e., escape or terminate) the ne gative situation, retali ate against those who caused the situation, and/or cope with th e negative stimulus through the use of illicit substances (Agnew, 1992, p. 58). Although the three types of strain are th eoretically distinct, Agnew asserts that there may be overlap in the measurement of these types of strain (Agnew, 1992, p. 59). For example, insults from a parent could be operationalized as a measure of failure to achieve positively valued stimuli, removal of positive stimuli, and/or a presentation of negative stimuli. Regardless of how a nega tive relation or cond ition is classified, each experience of strain increases the likelihood that one or more negative emotions will be felt. Moreover, Agnew asserts that strain may have a cumulative effect on individuals (Agnew, 1992, pp. 62-64) such that a person expe riencing one item of strain will be less affected than a person experiencing numerous items of strain. This assertion has been interpreted by many researcher s to advocate the use of a “c umulative” index of strain when operationalizing measures of various strainful events and conditions (Agnew, 2001, p. 324). However, Agnew suggests that differe nt types of strain (failure to achieve positive goals, removal of positively valued stimuli, and presentation of negative stimuli) will impact delinquency differently (p. 324). In studies of GST that have examined separate measures of strain (e.g., Ag new & Brezina, 1999; Agnew & White, 1992; Aseltine et al., 2000; Paternoster & Mazerolle, 1994) some measures have been significantly related to delinquency, while othe r have not. Furthermore, the amount of

PAGE 39

28 variance explained by the various measures of strain in one particular model will vary, with certain measures explaining two, three, or more times the variance in delinquency. Agnew suggests that future tests of GST shoul d attempt to include separate measures of strain, rather than composite indices. (Since tests of GST do not uti lize a standard set of measures to assess strain, comparisons of stra in measures between studies are difficult, if not impossible.) Recently, Agnew has offered further specification of the types of strain most likely to lead to delinquency (Agnew, 2001). In this theoretical elaboration of GST, Agnew (p. 320) specifies that strain may be conceptualized in either objective or subjective terms. Objective strain refers to conditions or events that are disapproved by a social consensus, such as abuse, death, homelessness, and starvation. Subjective strain refers to conditions or events that are disa pproved on a more relative or individual basis, but not necessarily by the majority of society. Subjective strain may be more influential to delinquent outcomes (p. 322). The majority of GST research examines objective strain (p. 321), employing measures of strain that the overall society would id entify. If research relies on objective strain measures alone, stra in may be underestimated within samples. Agnew (2001, pp.320-322) emphasizes the need to examine both objective and subjective strain when testing GST, particularly when considering group differe nces in perceptions of strain.3 In addition, research from the stress literature indicate s that individuals may be subjective even in their appraisal of obj ective forms of strain (Agnew, 2001, p. 321). That is, individuals may agree that a list of obj ectively defined strains is indeed what they 3 The majority of GST literature to date has been limit ed to measures of objective strain. However, there are some exceptions (see Agnew & White, 1992; Baron & Hartnagel, 2002; Hay, 2003).

PAGE 40

29 would characterize as strain measures. Howe ver, they may disagree about the strength or degree of each objective strain measure contained on the li st. Hence, individuals may be subjective in their evaluation of objective strain. These subjec tive appraisals of objective strain may be affected by various factors bo th internal (e.g., pers onality traits, selfefficacy, self-esteem, values/goals) and external (e.g., social support, life circumstances) to the individual (see Dohrenwend, 1998, 2000; Kaplan, 1996; Lazarus, 1999). Furthermore, the subjectivity or degree of magnitude for objective strain measures may change in over time, such that what one view s as highly strainful at one cross-section in time may become less strainful or not strainfu l at all at another pe riod in time. Agnew (2001, p. 322) suggests that examination of chan ges in the subjectivity of objective strain measures may lead to bette r understanding of the dynamics of the strain-delinquency relationship, especially the role that ne gative affect plays in this relationship. In addition to examining the influence of subjective measures of strain, Agnew (1992, pp. 64-66) asserts that strainful conditi ons and events may be more criminogenic when they are greater in magnitude (i.e., more problematic for the individual), recent greater in duration (i.e., chronic strain – see Whea ton, 1994; Turner, Wheaton, & Lloyd, 1995), and/or closely clustered temporally All else being equal, individuals who experience strainful conditions or events th at are more problematic (i.e., magnitude— conceptually similar to subjectivity), will be more likely to experience negative affect and cope through delinquency than those perceiving such strain as less problematic (Agnew, 1992, pp. 64-65). All else being equal, recent st rainful events will be more consequential to negative affect and delinquent behavior, than those that occurred some time ago (p. 65). GST research has not yet identified what the appropriate lag be tween strain and the

PAGE 41

30 expression of negative affect and delinquenc y is; nor has it iden tified how much time must lapse after strain for there to be no effects on negative affect or delinquency. All else equal, individuals who experience a strainful event over a long period of time (chronic) will be more likely to experience negative affect and delinquency, than those not experiencing long durations of strain (p. 65). Finally, all else equal, individuals experiencing several strainful events clustered closely in time will be more likely to feel negative affect and respond w ith delinquency, than those not experiencing clustered strain (pp. 65-66). In his 1992 explication of the theoretical foundation of GST, Agnew (1992) suggested that complete tests of GST should examin e the impact of these four influential factors on the strain-neg ative affect-delinquency relationship. Since his original publication of GS T, Agnew (2001) has specified four alternative factors that influence delinquency. Strainful conditions and events are more likely to lead to crime when they are charac terized as (1) unjust, (2) high in magnitude, (3) associated with low social control, a nd (4) associated with exposure to delinquent peers, their beliefs, and their approval (derived from social learning and r outine activities theories) (pp. 326-342). According to Agnew, the literature indi cates a link between “unjust treatment and anger” (p. 327). Assu ming this is true and all else equal, individuals that experience high frequencies of “unjust” strain will be more likely than their counterparts to express ne gative affect in the form of anger, which increases the likelihood that delinquency will result (pp. 327-328) As mentioned above, strain that is high in magnitude is expected to increas e the likelihood of nega tive outcomes. The measurement of the magnitude of strain requir es measuring the subjec tivity of strain (pp. 332-333). Strain that is associated with low social control is also hypothesized to

PAGE 42

31 increase the likelihood of stra in leading to delinquency (pp. 335-336). According to Agnew, low social control may reduce the costs associated with delinquency and the availability of legitimate coping mechanisms fo r strained individuals (p. 335). Finally, strain that is associated with exposure to delinquent peer association and beliefs is hypothesized to be more criminogenic for indi viduals experiencing strain, compared to those not experiencing stra in (pp. 336-337). According to Agnew, exposure to delinquent peers and their subcultural beliefs increases the attract ion to illegitimate coping mechanism (pp. 336-337), thereby affect ing delinquent responses to strain. Agnew states that “…all four of these characteristics are r oughly equal in importance and that the absence of any one characteristic s ubstantially reduces the likelihood that strain will result in crime…” (p. 338). In an attempt to further clarify which st rainful conditions are most criminogenic, Agnew (2001, pp. 343-347) provides the following list of strainful conditions most likely to lead to delinquency/crime: (1) failure to achieve unconventional goals that are most accessible through crime (e.g., money, exciteme nt, status), (2) lack of parental attachment/bonding, (3) unpredictable and seve re parental discipline, (4) abuse and neglect, (5) negative school experiences, (6) low status em ployment (i.e., the secondary labor market), (7) homelessness, (8) peer a buse, (9) criminal victimization, and (10) prejudice and discrimination. At first glance some of these factors, such as low status employment and homelessness, may not seem in accordance with the definition set forth by Agnew for strain (i.e., negative relations with others). Yet, it is important to remember that strain results from the frustr ation or pressure produced by the inability to achieve or maintain positive goals/stimuli and block negative stimuli. In most cases,

PAGE 43

32 strain is perceived as or a ttributed to a consequence of human interaction. In other words, people prevent other people from achie ving or maintaining positive stimuli and blocking negative stimuli. These consequences may be caused by individuals (such as an employer, a potential employer, or an abusiv e parent) or by society in general (such as society’s apprehension toward providing certa in individuals with gainful employment [i.e., labor market problems; s ee Baron & Hartnagel, 2002]). Negative Affect As briefly mentioned earlier, Agnew posits that an individual will experience at least one negative emotion, called negative affect per experience of stra in. He states that negative affect may cover a broad range of emotions, including depression, anxiety, despair, and grief, but the most influen tial of these emotions is anger (Agnew, 1992, 2001). Anger is important for general strain theory because it is one of the most potent reactive emotions. For instance, anger has been shown to affect the development of legitimate coping mechanisms by hindering abi lities to effectively express grievances, preventing recognition of suita ble styles of conflict re solution (see Colvin, 2000), interfering with perceptions of the costs of illegitimate responses, and fostering desires for revenge or retribution (Averill, 1982, 1993; Bernard, 19 90; Tedeschi & Felson, 1994; Tedeschi & Nesler, 1993; Tyler et al., 1997; Zillman, 1979). Re search has also indicated that anger is associated with unjust/inequ itable treatment (Agnew, 1992; Averill, 1982, 1993; Berkowitz, 1993; Tedeschi & Felson, 1994; Tedeschi & Ne sler, 1993; Tyler et al., 1997). Moreover, some studies have indicated that anger may significantly affect crime, especially acts of viol ence (Aseltine et al., 2000; Berkowitz, 1993; Brezina, 1998;

PAGE 44

33 Mazerolle et al., 2000; Mazerolle & Piquero, 1998; Piquero & Sealock, 2000; Tedeschi & Felson, 1994). Coping Mechanisms Whether or not negative a ffect leads to an illegitimate response depends on the availability of individual copi ng strategies. The adaptation of certain coping strategies may lead to deviant responses to strain, while others may prevent deviance. Agnew (1992, pp. 66-70) presents three classifications of coping strategies: cognitive, emotional, and behavioral. Cognitive coping strategies refer to the interna lization of strain such that is relates to the individuals’ goals, beliefs, values, and/or identity (p. 67). Agnew states that cognitive coping strategies include the employment of neutralization (e.g., “It doesn’t matter.”) and minimalization techniques (e.g., “It could be worse.”) in an effort to make strain seem nonexistent, less important, or somewhat deserved (pp. 66-69). Emotional coping strategies also refer to the internal ization of strain, but they pertain to the emotional, rather than “rational,” state of the individual only ( pp. 69). According to GST, emotional coping strategies include both legitimate and illegitimate acts, such as physical exercise, meditation, and illicit and lic it substance use, which are utilized to reduce negative affect (pp. 69-70). Behavior al coping strategies refer to external responses to strain (p. 69). Behavioral coping strategies include attempts to reduce or eliminate sources of strain (e.g., regain posit ive valued stimuli when they have been blocked or lost and attempts to block or te rminate the source of negative stimuli) and attempts to seek revenge against those cau sing the strainful cond itions (p. 69). Agnew acknowledges that in the list of coping mech anisms included in his foundation of GST is not a complete list, but suggests they are the most prominent (p. 70).

PAGE 45

34 GST posits that individuals may choose from several forms of legitimate and illegitimate coping strategies when confronted with strain. Indi viduals who experience high levels of strain and choose illegitimate coping mechanisms will be more likely to respond to strain with deviance. On its face this statement seems conceptually tautological; however, severa l constraints and other condi tioning factors may prevent illegitimate coping strategies from leading to deviance (pp. 70-74). Although individuals have a choice in which coping mechanisms to us e, this does not imply that the choice is a free, rational, or conscious decision. Fu rther, coping strategies are not equally distributed. Different indi viduals will have access to different coping strategies depending upon a variety of other cond itioning and dispositional factors (e.g., temperament, self-esteem, so cial support) (pp. 70-74). Other Conditioning Factors Agnew suggests that the presence of certa in coping strategies is not the only factor conditioning whether or not an indivi dual will choose illegitimate responses to strainful conditions. Agnew describes a rather extensive, yet partial, list of internal and external factors that may further influen ce the effects of stra in (Agnew, 1992, pp.70-74). Internal factors include such characteristic s as temperament, intelligence, and beliefs; external factors refer to char acteristics such as structura l/environmental circumstances and existing social support structures. Ma ny of these conditioning factors have been incorporated in Agnew’s 2001 explication of the 10 most strainful conditions most likely to lead to delinquency/crim e (discussed above). GST assumes that the presence and/or absence of certain conditioning factors encourage problem-solving and act as buffe rs against strainful situations, thus

PAGE 46

35 ameliorating much of the negative effects of strainful situations. Individuals who possess beliefs, goals, and value definitions in line with conventiona l society, agreeable temperaments, higher intelligence, interper sonal skills, higher self-esteem, self-efficacy, problem-solving skills, conventional social support systems, a learning history reinforcing conventional behavior, dispositiona l attributions of blam e, and environmental characteristics that are socially organized a nd lack subculturally deviant influences will be more likely to select non-delinquent c oping strategies, than those who lack these traits/conditions (Agnew, 1992, pp. 70-74). Conditioning factors may also influence the level of subjectivity for strain. That is, internal and external condi tioning factors not only modera te the relationship between strain and coping mechanisms, but also the degr ee to which individuals perceive strain as problematic and negative a ffect-inducing (Agnew, 2001, p. 333). This suggests internal and external characteristics ma y directly influence individual levels of strain. Although GST has not fully conceptualized how these conditioning factors may operate, it does suggest that there are multiple factors grounded within various scientific paradigms that are associated with and even cause strain. GST does not fully explicate how conditio ning factors affect illegitimate coping strategies to strain and ne gative affect, nor does it speci fy which factors are more important than others. However, it seems logi cal to assume that individual characteristics such as personality, morals, intelligence, and confidence (i.e., self-esteem, self-efficacy) will affect how individuals react to their en vironment. For example, individuals with high confidence may be more likely to cope w ith negative relationships with others by convincing themselves that the strain does not matter—that they can achieve their goals

PAGE 47

36 regardless of the opposition. In another exampl e, individuals that experience negative peer relationships such as being bullied or teased at school who have other positive social support networks in place (e.g., church and family ) may be better able to cope with strain through legitimate strategies because the pres ence of these prosocial support systems helps to alleviate some of th e effects of strain. Whereas, individuals who experience the same bullying at school, but lack prosocial support, may turn to drug use or delinquent peer associations to re lieve the pressures of the strain. These are just a few examples. Certainly, there are many possible internal and external resources that can influence whether or not strain leads to delinquency. The present study is particularly con cerned with how pers onality traits or temperament influence the strain-delinquency re lationship. In the first and only test to date of GST and personality traits, Agnew et al. (2002, p. 45) have stated that “…the impact of such [personality] traits may be fa r more pervasive than that of the conditioning variables typically examined…” in the GS T research. They have suggested that personality traits can influence individual emotionally responds to strain and the development of deviant coping strategies (Agne w et al., 2002, p. 45). Part of the impetus for suggesting a test of GST and personality wa s derived from a theoretical discussion of the application of GST in the explanation of differences in life-c ourse trajectories of crime (Agnew, 1997). In this paper, Agnew s uggests that personality traits may influence why certain individuals stop offending after adolescence and other continue to offend throughout their lifetime (see Moffitt, 1993). Personality traits describe one’s perception and behavior toward the environment and are relatively stable charac teristics that are to a certain degree inherent in nature (see

PAGE 48

37 Caspi & Bem, 1990; Ge & Conger, 1999; Roberts & DelVecchio, 2000; Soldz & Vaillant, 1999). Personality traits have been measured in a variety of ways (see John & Srivastava, 1999). Traits may be describe d by numerous facets of interpersonal, affective, and behavioral terms and definitions such as whether an individual is sociable, warm, trustworthy, modest, sympathetic, orga nized, responsible, laz y, impulsive, hostile, anxious, content, imaginative, and so on a nd so forth (see John & Srivastava, 1999). Personality traits can be both conforming ( normative) and non-conforming (maladaptive) to society. As such, individua ls possessing traits that may make them less able to control emotional responses and impulsivity and mo re inclined to ex press and experience negative emotions such as ange r, anxiety, and fear, particular ly under stressful situations will be more likely to cope with strain through illegitimate coping mechanisms (Agnew et al., 2002). Figure 1 presents a full models of Agne w’s (1992, 2001) general strain theory. Each text box represents a key concept of GST. Path relationships are represented by solid arrows, pointing in the cau sal direction of the relations hip. Mediating relationships are represented by dotted lines, connected at a po int on the path its associated measure is hypothesized to mediate. The three main types of strain are illustrated in the model: (1) failure to achieve positive goa ls, (2) removal of positive stimu li, and (3) pr esentation of negative stimuli. Strain leads directly to delinqu ency (refers to any illi cit activity). Strain also leads indirectly to delinque ncy via negative affect. Negativ e affect is directly related to delinquency. Internal and external c onditioning factors are hypothesized to lead directly to strain and mediat e the direct and indirect effe cts of strain on delinquency. Behavioral, cognitive, and emoti onal coping strategies are predicted to mediate the direct

PAGE 49

38 and indirect effects of stra in on delinquency. Of course this model is a simplified representation of the complex re lationships presented in GST. The mediating effects of conditioning factors and coping strategies may not be appropriate for all measures that included within these concepts.

PAGE 50

39 Figure 1: A Model of General Strain Theory Conditioning Factors (e.g., beliefs/values, temperament, intelligence, selfesteem, self-efficacy, interpersonal skills, social support, envi r onment ) Failure to Achieve Positive Goals (aspirations vs. expectations; expectations vs. achievements; just outcomes vs. actual outcomes) Removal of Positive Stimuli Presentation of Negative Stimuli Delinquency Negative Affect Coping Mechanisms (behavioral, cognitive, & emotional)

PAGE 51

40 Empirical Support for General Strain Theory Strain-delinquency. Several studies have provide d empirical support for the propositions Agnew has set forth in GST. A significant positive relationship between various strain measures and delinquency ha s consistently been reported (Agnew, 1985, 1989, 2002; Agnew & Brezina, 1997; Agnew et al., 2002; Agnew & White, 1992; Aseltine et al., 2000; Bao, Haas, & Pi, 2004; Baron & Hartnagel, 1997, 2002; Benda & Corwyn, 2002; Benson, Fox, DeMaris, & Va n Wyk, 2003; Brezina, 1999; Broidy, 2001; Eitle, 2002; Eitle & Turner, 2002, 2003; Hoffmann, 2002; Hoffmann & Cerbone, 1999; Hoffmann & Ireland, 2004; Hoffmann & Mi ller, 1998; Hoffmann & Su, 1997; Kim, Conger, Elder, & Lorenz, 2003; Maxwell, 2001; Mazerolle, 1998; Mazerolle et al., 2000; Mazerolle & Maahs, 2000; Mazerolle & Piquero, 1997, 1998; Mazerolle, Piquero, & Capowich, 2003; Paternoster & Mazerolle, 1994; Peter, LaGrange, & Silverman, 2003; Piquero & Sealock, 2000, 2004; Robbers, 2004; Sharp, Brewster, & Love, 2005; Sigfusdottir, Farkas, & Silver, 2004; Wallace, Patchin, & May, 2005; Warner & Fowler, 2003). For example, negative life events have been consistently re ported as related to delinquency. Agnew (2002) has even found that certain forms of vicarious strain are significantly related to delinque ncy, especially experienced victimization and vicarious victimization of family and friends. Studies have also indicated that alco hol and illicit drug use may reduce the experience of stress (e.g., Conrod, Pihl, & Vassileva, 1998; Newcomb, Chou, Bentler, & Huba, 1988; Sayette, 1993). A significant posit ive relationship between various strain measures and substance use has also been noted across a considerable body of literature (Agnew & White, 1992; Aseltine et al., 2000; Boardman, Finch, Ellison, Williams, &

PAGE 52

41 Jackson, 2001; Eitle, 2002; Hoffmann & Su, 1997; Peter et al., 2003) Other empirical studies have shown both delinquency and subs tance/alcohol abuse, combined, to be positively related to strain (Agnew & Br ezina, 1997; Agnew et al., 2002). Mediating influence of negative affect. On the other hand, empirical studies of the indirect relationship between strain and delinquency, when mediated by negative affect, have been less consistent. These findings ma y be due to the use of varying measures across studies. While some GST researchers have used composite measures of negative affect (i.e., combining experiences of anger, anxiety, depression, etc. into one index), others have examined negative emotions se parately. Thus, inasmuch as GST scholars have rarely used the same combination of em otions in their studies, the comparability of findings across research has been hindered. Although strain has been significantly a nd positively associated with anger (Agnew, 1985; Aseltine et al., 2000; Bao et al., 2004; Brezina, 1996, 1998; Broidy, 2001; Hay, 2003; Jang & Johnson, 2003; Mazerolle et al., 2003; Mazerolle & Piquero, 1997, 1998; Piquero & Sealock, 2000, 2004; Sharp et al., 2005; Sigfusdottir et al., 2004), the direction and role of anger as a mediating variable on certain t ypes of delinquency is unclear. Some studies, however, appear to s upport the assumption that anger serves as a mediator between strain and both general and specific types of delinquency (Agnew, 1985; Aseltine et al., 2000; Bao et al., 2004; Brezina, 1998; Hay, 2003; Jang & Johnson, 2003; Sharp et al., 2005; Sigf usdottir et al., 2004). Fo r example, Jang and Johnson (2003) found that a measure of negative aff ect, including both internally-directed and externally-directed emotions, completely medi ated the effects of a composite measure of strain on measures of general devi ance, drug use, and fighting.

PAGE 53

42 Other findings have suggested that anger ma y be limited in its role as a mediator for the strain-delinquency relationship to measures of violence or interpersonal aggression, but not to acts of non-violent behavior (e.g., prop erty crimes) or substance use (see Aseltine et al., 2000; Piquero & Sealock, 2000). Even more perplexing, Mazerolle and associates (2000) de monstrated that it is actually strain that mediates the relationship between anger and violent delinque ncy. Along these same lines, Kim et al. (2003) have reported that internalizing pr oblems such as depression and anxiety may exacerbate stressful life conditions. A nother study conducted by Mazerolle and colleagues (2003) suggested that differences in the types of anger (i.e ., situational versus trait) may explain some of the inconsistencies regarding the role of negative affect. That is, trait anger was significantly related to violent forms of delinquency, while situational anger was shown to be signifi cantly related to both non-vi olent and violent forms of delinquency (cf. Capowich, Mazerolle, & Piquero, 2001). Yet, some studies have supported the assu mptions of general st rain theory when examining alternative measures of negative affect, such as composite measures of negative emotions (Broidy, 2001; Capowich et al., 2001; Sharp et al., 2005), anxiety (Bao et al., 2004; Brezina, 1996; Ki m et al., 2003; but see Aselti ne et al., 2000), depression (Bao et al., 2004; Brezina, 1996; Hagan & Fo ster, 2003; Kim et al., 2003; Piquero & Sealock, 2000, 2004), frustration (i.e., a mild form of anger) (Wallace et al., 2005), resentment (Bao et al., 2004; Brezina, 1996), and guilt (Hay, 2003). Even such alternative measures of nega tive affect have produced mixed results, however. For instance, in a study of conditioning factors of the strain-delinquency relationship among

PAGE 54

43 three waves of data for high school students, Aseltine et al. (2000) found no support of a significant mediating effect of anxi ety between strain and delinquency. Interestingly, Hagan and Foster (2003) reported that anger was actually a source of depression, particularly among females, and that the relationship between anger and depression is partially mediated by delinque ncy. Sharp and colleagues (Sharp, TerlingWatt, Atkins, & Gilliam, 2001) have also re ported a connection between depression and anger. In a test of general strain theo ry on purging behaviors among college women, the authors indicated a moderating effect of de pression on the relationship between anger and purging. In contrast, when depression wa s high, anger was significantly and positively related to purging, but when depression wa s low, anger had no significant effect on purging. Furthermore, using structural e quation modeling, Sigfus dottir et al. (2004) found that significant mediating effects of depression on the strain-delinquency relationship are suppressed when controlling for anger. Studies have also indicated that delinquency reduces the impact of strain on negative affect (Brezina, 1996; Hoffman & Ireland, 2004), though this moderating effect appear s to be more important with regard to anger than other forms of negative affect. Moderation/mediation of copi ng mechanisms and other factors. According to general strain theory, certain coping strate gies and conditioning factors are hypothesized to reduce (when conducive to legitimate behavior) or enhance (when conducive to illegitimate behavior) the effect s of strain on delinquency. However, research has lacked empirical consistency with respect to ex amining those forms of individual coping strategies posited to di rectly affect how the individual ad apts to strain. When combined, these studies include several measures of conditioning factors such as self control, self-

PAGE 55

44 esteem, self-efficacy, delinquent peers, fam ily communication, moral beliefs, religiosity, and social support. Several studies have provided empirical support to claims made by GST related to self-efficacy (Agnew & White, 1992; Paternos ter & Mazerolle, 1994), delinquent peers (Agnew, 2002; Agnew & Brezina, 1997; Agne w & White, 1992; Bao et al., 2004; Baron & Hartnagel, 2002; Benda & Corwyn, 2002; Hay, 2003; Mazerolle et al., 2000; Mazerolle & Maahs, 2000; Mazerolle & Pi quero, 1998; Peter et al., 2003), delinquent norms or beliefs (Bao et al., 2004; Bar on & Hartnagel, 2002; Benda & Corwyn, 2002; Brezina, 1998; Hoffmann, 2002; Mazerolle & Maahs, 2000; Mazerolle & Piquero, 1998), and external attribution of blame (Baron & Hartnagel, 2002). Another body of research has focused upon GST concepts of self-esteem or self-concept (Benda & Corwyn, 2002; Hoffman & Ireland, 2004; Jang & Johnson, 2003; Piquero & Sealock, 2000), social support (Boardman et al., 2001; Robbers, 2004; Warner & Fowler, 2003), spirituality or religiosity (Jang & Johnson, 2003; Pique ro & Sealock, 2000), and community characteristics (e.g., unemployment) (Hoffm an, 2002). Moreover, conditioning factors have also been shown to influence negative affect (e.g., anger, re sentment, anxiety, and depression) as predicted by general strain theory (see Brezina, 1996; Jang & Johnson, 2003). Other researchers have reported conflicti ng results regarding the role certain conditioning factors play in the moderation of strain (Aseltine et al., 2000; Boardman et al., 2001; Capowich et al., 2001; Eitle & Turner, 2003; Hoffmann & Cerbone, 1999; Hoffmann & Miller, 1998; Piquero & Sealoc k, 2000). For example, in a three-year longitudinal analysis of the conditioning e ffects of self-efficacy and self-esteem on the

PAGE 56

45 strain-delinquency relationship, Hoffma nn and Miller (1998) found no support for Agnew’s claims that self-efficacy and self-est eem moderate the effects of negative life events on delinquency. That is, their analyses revealed that youths who did not associate with delinquent peers were significantly more lik ely to report increased strain that led to a rise in delinquency. In a study of genera l strain theory among a juvenile offender population, Piquero and Sealock (2000) examin ed the moderating effects of five coping skills (cognitive, emotional, social, physical, and spiritua l) on two forms of negative affect (anger and depression). Their study found only margina lly significant effects for interaction terms between depression and em otional coping skills and between depression and spiritual coping skills. Similar to delinquency, the stress literature has indicated that conditioning factors and coping mechanisms affect the relationshi ps between stress and substance use (e.g., Brook, Nomaura, & Cohen, 1989; Carvajal, Cl air, Nash, & Evans, 1998; Weinrich, Hardin, Weinrich, & Valois, 1997). Moderati ng effects of conditioning variables on the strain-substance use relationship ha ve been reported in the general strain literature as well (see Agnew & White, 1992; Jang & Johnson, 2003). Clearly this review indicates the need to better understand th e salience of coping on the development of antisocial behaviors and emotions. Distinguishing General Strain Theory from Social Control and Social Learning Theories Three of the leading cr iminological explanations of general delinquency are strain, social control, and social le arning theories (Agnew, 1992, 2001; Agnew & Brezina, 1997; Alarid, Burton, & Cullen, 2000; Battin, Hill, Abbott, Catalano, & Hawkins, 1998; Bernburg & Thorlindsson, 1999; Kornhauser, 1978). Other theoretical

PAGE 57

46 explanations of crime have been developed; however, these three rubrics have emerged as the dominant micro-level theories in main stream criminology. In his discussions of the theoretical assumptions of GST, A gnew (1992, 2001) emphasizes the importance of including measures of both so cial control and social lear ning/differential association theories to test the empiri cal validity of GST. In a recent explication of the specific types of strain most likely to cause delinquency/crime, and toward this end, two of the four characteristics of the most criminogenic strainful conditions were derived exclusively from social control and social learning theories (Agnew, 2001, pp. 335-338). First, strain that is caused by or associated with low social control (derived from social control theory) will be more likely to lead to delinquency than strain that is not. Agnew (p. 335) contends that certain forms of strain, such as parental conflicts and parental reje ction, are caused by low social control. Low social control may also reduce access to le gitimate coping strategies. Therefore, individuals experiencing strain that is either caused by or associated with low social control may choose or have no other choice th an to choose illegitimate coping strategies, which may lead to delinquency. Second, strain that is associated with peer pressure/incentive to engage in criminal activity (derived in part from differential a ssociation/social learni ng theory) may be more likely to lead to delinquency than strain that is not (Agnew, 2001, p. 337). Individuals who are associated with delinquent subcultu res (e.g., gangs) or delinquent peers may be more likely to choose illegitimate coping strategies. In some instances, this may be due to partial or full assimilati on of delinquent beliefs and nor ms and/or reinforcement for modeling delinquent behaviors. In other cases certain types of strain may require that

PAGE 58

47 individuals respond with delinque ncy. For example, youths associated with gangs may be required to respond to disres pectful treatment from other youths with violence. Yet, when other forms of strain are experi enced, these youths may not respond with delinquency. Under the aforementioned circum stances, the conceptua lization of coping strategies appears to be pr edicated on social control and differential association. Social Control Theory Control theory (Durkheim, 1897/1951; Hirs chi, 1969; Nye, 1958) assumes that humans are inherently hedonistic, thus people are naturally inclined to violate rules. According to control theory, individuals conf orm to society’s laws because of social controls that prevent them from committing crim es. In other words, natural urges such as intimidation, retaliation, and vengeful retr ibution are controlled via bonding to conventional others and institutions (Hirsc hi, 1969). According to Hirschi’s social bonding theory (Hirschi, 1969), (1) the level of attachment to parents, teachers/school, peers, or other institutions (e.g., church), (2) commitment (actua l or anticipated) in conventional society, (3) involvement in conve ntional activities, and (4) internalized conventional beliefs an individual possesses are all inversely related to deviance. Deviance results only when these social cont rols are weakened or broken (Hirschi, 1969; Reiss, 1951). Thus, individuals lacking positiv e social controls, whether long-term or episodic (see “drift” theory: Matza, 1964), are therefore “free” to satisfy their needs by utilizing delinquent means. Another form of control theory that has received attention in the GST literature is Gottfredson and Hirschi’s (1990) self-control theory. In cont rast to social bonding theory (Hirschi, 1969), whereby self-c ontrol is a subsumed concept of attachment, this new

PAGE 59

48 theoretical framework revolves completely aroun d the concept of self -control. In their general theory of crime, Gottfredson and Hirschi (1990) state the following: [P]eople who lack self-contro l will tend to be impulsive, insensitive, physical (as opposed to mental), risk-taking, shortsighted, and nonverbal, and they will tend therefore to engage in criminal and an alogous acts. Since these traits can be identified prior to the age of responsibility for crime, since there is considerable tendency for these traits to come together in the same people, and since the traits tend to persist through life, it seems reasona ble to consider them as comprising a stable construct useful in the explanation of crime. (pp. 90-91) Although individuals with low self-control will be more likely to commit criminal acts, not all individuals low in self-control will commit crimes. Whether or not an individual will become criminal depends on the existence of several intervening mechanisms, such as parenting style, parent al attachment, punishment for deviance, and parental recognition of devian ce (related to values/beliefs). Many of these intervening mechanisms are those described by the four elements of social bonding theory, though more proximal in nature. Therefore, as Ak ers (1997, p. 92) succinctly states, “It may be then assumed that self-control is the key variable, and that ot her social bonds affect crime only indirectly through their e ffects on self-control.” Thus Agnew’s specification that tests of GST include strains associated with lo w social control calls for measures of selfcontrol, social bonding, or both. Strain will be more likely lead to delinquency if these social controls are sufficiently weakened.

PAGE 60

49 Social Learning/Differential Association Theory In general, the phrase “social learning theory” refers to any social behavioral explanation. In the field of psychology, social learning theory refers to the “…reciprocal interaction between cognitive, behavioral a nd environmental determinants…” (Bandura, 1977, p. vii), and includes research by Bandur a and other psychologists (Bandura, 1977; Bandura & Walters, 1963; Rotter, 1954). Hist orically, psychologist s and sociologists have applied the concepts of social learning to examinations of de viance and delinquency (Jessor & Jessor, 1977; Patters on, 1995; Patterson & Chamberl ain, 1994; Patterson, Reid, & Dishion, 1992; Patterson, Reid, Jones, & C onger, 1975). When criminologists refer to social learning theory, however, this usually pe rtains to the theory as it was developed by Ronald Akers (Akers, 1973, 1977, 1985; see als o, Burgess & Akers, 1966) as a revision of Sutherland’s differential associ ation theory (Sutherland, 1947). Unlike social control theory, social lear ning theory does not assume that humans are inherently deviant creatures, but are rather creatures that learn deviant behavior from others. Akers’ social learni ng theory offers an explanati on of deviance that describes processes that function both to motivate and control deviant behavior, thus serving both to undermine and promote social conformity. Akers explicates the following four primary processes whereby delinquent beha vior is learned (Akers, 1973, 1977, 1985): (1) regular association with others who engage in deviant acts– differential association (2) anticipated and actual rewards reinforcing delinquent behavior outweigh the costs of deviance– differential reinforcement (3) imitation or modeling delinquent behavior after observation of such behavior being committed by others– imitation and/or (4) transmission of delinquent attitudes and values– definitions

PAGE 61

50 From a social learning perspective, the motive for delinquency is the anticipated rewards net other costs (i.e., reinforcement) fo r such behavior. That is, individuals who observe others commit delinquent acts that are perceived to result in more positive outcomes than negative outcomes are more lik ely to imitate this behavior. When the results of their imitation are also more positiv e than negative, they are more likely to adjust their own moral attitudes and beliefs to condone such behavior Individuals with increased exposure to deviant ot hers (i.e., high differential asso ciation) are more likely to anticipate reinforcement for deviance and are thus more susceptible to delinquent behavior. According to Agnew (1992, 2001), associ ations with deviant others may condition the effects of strain on delinquency by increasing the appeal and availability of illegitimate coping mechanisms or limiting leg itimate coping mechanisms. For example, certain types of strain, such as abuse fr om parents and/or peers, may increase the likelihood that youths will asso ciate with delinquent others (Agnew, 2001, p. 337). This exposure increases the likelihood that they wi ll internalize delinquent beliefs and values and witness positive or reinforcing consequen ces for delinquent behavior. Therefore, delinquent peer association may influence a youth’s perceived costs and rewards for employing delinquent coping strategies for st rain. In another example, individuals belonging to certain subcultures (e.g., juveni le gangs) may be “bound” by the definitions and beliefs of that subculture to respond to certain types of strain, such as disrespect or negative relations with other subculture me mbers, with delinquency (Anderson, 1999). These same individuals, however, may choos e legitimate coping strategies for other

PAGE 62

51 sources of strain. As a result of this r easoning Agnew asserts that tests of GST should include or control for measures of differential association. Two Weaknesses of Tests of GST: Tautology and Falsifiability Despite Agnew’s emphasis on the need to control for social control and differential association in tests of GST, it is often the case that measures of strain are correlated with measures of social control and social le arning theory (Agnew, 2001); consequently, empirical analyses that m odel strain along with control or learning variables may produce, in effect, empirical taut ologies. In these studies, the theoretical constructs being measured for strain (e.g., strict parental discipline, low grades in school) may overlap with constructs of social contro l theory and differentia l association/social learning theory (Agnew, 1995). That is, a cons truct being measured and tested for strain theory may also be used to test either soci al control or differential association without changing its operationalization. In an effort to minimize this issue, some researchers have suggested that one solution for overlapping theoretic al concepts depends on the di stinctions made about the construct when it is being tested within a particular theoretical model. That is, researchers may simple a priori clarify to which constructs certain measurement items will be assigned. To quote Hay (2003: 118): “The challenge in testing GST is to identify social control and social learning variables that cannot reasonably be seen as a strain theory variable.” Unfortunately, as A gnew notes (Agnew, 2001, pp. 348-350), this task often proves difficult to achieve. Some re searchers have attempted to maintain theoretical distinction by “… assigning some measures to the strain camp, some to the social control camp, and some to the social-learning camp.” (Agnew, 2001, p. 348). If

PAGE 63

52 strong associations exist after delineating wh ich overlapping constr ucts will be assigned to each theory, how does one know if the stat istical significance achieved by the strain measure is truly a result of “strain” rather th an social control or social learning (Agnew, 2001, p. 349)? Agnew (Agnew, 2001, pp. 349-350) offers three potential ways, critical tests, to rectify the issue regard ing overlapping constructs. First, tests of GST could consider the intervening processes described by each of the theories. The three theories differ in their explanation of how and why (i .e., intervening processes) de linquent/criminal behavior results. General strain theory focuses on th e role of intervening mechanism of negative affect. Social control theory focuses on th e intervening mechanism of perceptions of lowered costs for delinquency and also a ssumes a direct, non-mediated, relationship between low social control and delinquency. Social learning theory focuses on the role of the intervening mechanism of perceptions of desirability (beliefs and reinforcements) of delinquency. Therefore, empirical studies of overlapping parental, peer, school, and work concepts including these interventions sh ould be able to distinguish which theory best explains the causal relationship between ove rlapping strain, social control, and social learning measures and delinquency. A critical test of GST that includes an examination of the intervening mechanisms of GST, social control, and social learning theory is quite feasib le; however, most data sets do not contain enough of the relevant meas ures to conduct such s test. For the sake of argument, assuming one had a large, natio nally representative data set that included several measures of strain, social control, and so cial learning theories. In this critical test of GST, it would not be necessary to exam ine all types of strain, only those that

PAGE 64

53 conceptually overlap with social control (negative relationships with parents and teachers) and differential association (delinquent peer associations). On the other hand, it would be crucial to include measures of ne gative affect (especially anger), differential reinforcement, and deviant beliefs/values as me diators in the model. If analyses revealed that negative relationships with others (strai n) or low social control significantly led to delinquency through the mediation of negativ e affect, a GST perspective would be supported. If the results indica ted that negative relationships or low social control led directly, with no significant mediation of ne gative affect, to delinquency, then a social control approach would be supported, not GST. If the data indicated that delinquent peer associations led to delinquency when mediat ed by delinquent beliefs and/or differential reinforcement for delinquent behavior, then social learning theory would be supported. (It is also possible that lo w social control/strain measures may lead to delinquency through differential reinforcement, which would suggest either a social control or social learning explanation. However, if the intenti on of the test is to examine the empirical support for GST, such a finding would still sugge st that GST is not a viable explanation for delinquency.) Second, tests of GST could control for the effects of social control and social learning in the analysis of the effects of strain. Since Agnew argues that strain may affect delinquency by lowering social control a nd increasing incentives to engage in delinquency, statistically significant effects of strain on delinquency, when controlling for social control and social learning, would suppor t GST. This approach, however, is not applicable when utilizing measures of strain that “directly index” the more relevant measures of social control or social l earning (e.g., low grades) (Agnew, 2001, p. 349).

PAGE 65

54 Thirdly, Agnew recommends that te sts of GST could include neutral relationships, in addition to positive and negative relationshi ps, in the ope rationalization of strain, social control, a nd social learning measures. W ith respect to social control theory, neutral relationships (i.e., apathetic) and negative relationships with conventional others should lead to delinquency. Fr om a GST perspectiv e, however, neutral relationships with conventiona l others, particularly parent s, are neither a symptom nor source of strain. Therefore, neutral relationshi ps with others should not be criminogenic. In one of the only studies to examine competing predictions of GST and social control theory by examining posit ive, neutral, and negative re lationships with parents and teachers, Thaxton and Agnew (2004) used polynomial regression to examine the relationships between parental and teacher attachment and delinquency. They employed a graphic interpretation of the regression of delinquency on a semantic differential scale of attachment ranging from negative to neutral to positive to test whether social control theory or GST was a better predictor that low attachment leads to delinquency. Imagine a line graph, where the x-axis is attachment ranging in value from 1 to 10 with 1 being negative attachment, 5 being neutral attachment, and 10 being positive attachment, and the y-axis is delinquency ra nging from none to high. If the slope of the non-linear regression line was near or at zero when delinquency was highest and attachment was lowest (negative attachment), started becoming negative when attachment was 5 (neutral attachment), and approached zero again as delinquency approached 0 and attachment approached 10 (positive attachment), then a social control perspective would be supported by the data. On the other hand, if the slope of the regr ession line was negative when delinquency was highest and attachment was 1, started to quickly approach zero

PAGE 66

55 when delinquency was low and attachment was 5 (neutral attachment), and continued to flatten out as attachment approached 10 (pos itive attachment), a GST perspective would be supported by the data. Thaxton and Agne w (2004) reported the shape of the curve describing the relationship between attachment and delinquency supported GST. That is, negatively attached youths were substantially more delinquent than either neutral or positively attached youth, who were comparably delinquent. Despite these three strategies offere d by Agnew to control for overlapping theoretical constructs when testing GST, he admits that none will provide a “perfect” empirical determination of the e ffects of strain on delinquency. It is therefore crucial that tests of GST recognize such weaknesses and attempt to control for them wherever possible. In addition to issues with regard to measurement and testing, a second major weakness is that GST studies suffer from the s eemingly unfalsifiable na ture of the theory. Although general strain theory is a relatively new theory, it has receiv ed much theoretical elaboration over the past decade. These theoretical expansions have increased the scope of strain beyond that estab lished in the origin al foundation of the theory, which was already arguably broad. Consequently, the testability of GST has criticized (Jensen, 1995). Accord to Jensen, “If strain can be de fined in so many different ways, then strain theory is virtually unfalsifiable. There is always a new measure th at might salvage the theory.” (Jensen, 1995, p. 152). Perhaps the solu tion to proving the falsifiability of GST lies with improving the operationalization and st atistical techniques utilized to test the theory. If datasets contained appropriate measures to address Agnew’s aforementioned three strategies to separate e xplained variance attributed to strain theory measures from

PAGE 67

56 that of social control and social learning measures, tests may reveal some enlightening findings that could diminish such criticisms. It is evident that there are “chinks in the armor” of GST. On one hand, Agnew advocates GST as a complementary theory (Agnew, 1992, p. 76) to social control and social learning theories. He even goes as fa r as specifying that strain will be more criminogenic when characterized by associa tion with low social control and social learning mechanisms (i.e., delinquent peers). On the other hand, Agnew stresses the need to control for social control and social lear ning measures, thus treating the theories more like competing, rather than complementing th eories. Certainly this inconsistency has contributed to the harsh crit icisms by some social scie ntists by labeling GST as “unfalsifiable”. While not resolving thes e limitations, the present study treats social control and social learning theories as co mpeting, yet highly associated theories by including social contro l and social learning models in the models and co rrelating them with the strain measures. In addition, ever y effort will be made to ensure operational distinction among the measures. This chapter explained the key construc ts of GST and the importance of social control and differential associ ation in tests of GST. Agnew (1992, 2001) consistently stated that social control a nd differential association may c ondition the effects of strain on delinquency, but has only recently offered a critical test of GST and social control (Thaxton & Agnew, 2004). Agnew (1992, 2001) has also implicitly stated that personality may condition the strain-delinque ncy relationship, and recently provided a test of GST and the moderating effects of personality traits (Agne w et al., 2002). The purpose of this study was to elaborate on Agnew et al.’s test of personality and GST.

PAGE 68

57 However, before a discussion of the Agnew et al. (2002) article and the proposed study are presented, it is necessary to understand how personality relates to antisocial behavior. The next chapter presents a discussi on of normative (i.e., conforming) and maladaptive (i.e., non-conforming) personality tr aits. The relationship between traits and motives is discussed. Next, a descripti on of normative personality traits, including methods used to assess personality traits and empirical correla tes of personality traits, is presented. Then, maladaptive personality tr aits, specifically psychopathic personality traits or psychopathy, are de scribed. Finally, issues of tautology, when examining personality/psychopathy and delinquency, are mentioned.

PAGE 69

58 Chapter 3 Personality Characteristics Affecting Deviance As stated in the previous chapter, gene ral strain theory expanded the scope of sources of strain and explicat ed factors that affect the strain-crime relationship (e.g., negative affect, coping mechanisms, and other conditioning factors). Since the introduction of GST, Agnew and others have attempted to improve the comprehensiveness of the processes described in the assu mptions of strain theory. Often these theoretical expansions of ge neral strain theory have b een guided by the findings of previous studies. These embellishments have provided more specification of criminal motivations (Agnew, 1992), criminogenic types of strain (Agnew, 2001), gender differences (Broidy & Agnew, 1997), structural effects that may condition the straincrime relationship (Agnew, 1999), developmenta l or life-course diffe rences in strain (Agnew, 1997), and biological e xplanations of the strain-crime relationship (Walsh, 2000). Until recently, the GST literature has ignor ed what Agnew and associates (Agnew et al., 2002) argue may be one of the most important conditioning e ffects of the straincrime relationship, namely the personality trai ts of the individual experiencing strain. Personality traits are relativel y stable characteristics that describe one’s perception and behavior toward the environment (Caspi et al., 1994). There is impressive empirical evidence that suggests personality traits ma y be stable and enduring characteristics,

PAGE 70

59 which are affected by biological and early socialization processes. In his early postulation of GST, Agnew ( 1992) made no specific menti on of personality traits or psychopathic features. He did, however, make reference to internal coping mechanisms such as “temperament.” He further stated that chronic st rainful conditions may “…have a greater impact on a variety of negative ps ychological outcomes.” (Agnew, 1992, p. 65). This statement suggests that personality tr aits may influence the strain-delinquency relationship. Several years la ter, Agnew stated that “[t]he subjective evaluation of an objective strain is a function of a range of factors, including individual traits (e.g., irritability)…” (Agnew, 2001, p. 321). Hence, one can make a tenable argument that personality traits may serve as moderati ng and mediating factors for the straindelinquency relationship. Trait or Motive? Upon examination of the psychological l iterature, a novice to the field may ask what the difference is between a motive and a personality trait This question is particularly important when testing genera l strain theory, considering that Agnew (Agnew, 1992, 2001; Agnew et al., 2002) re peatedly emphasizes that GST is distinguished from other criminological theories because of its focus on motivational processes. Unfortunately, these concepts are not universally defined; there exists much overlap and ambiguity with regard to the mean ing of these two concepts (for a discussion see Winter, John, Stewart, Klohnen, & Dunca n, 1998). According to Winter et al. (1998), motives refer to one’s desires or goals that vary in relation to the situation or more immediate circumstances (c onceptually similar to strain). That is, as circumstances and situations change, motives ch ange, and are, therefore, part icularly unstable over time.

PAGE 71

60 Consequently, motives may be difficult to m easure either directly or through observation, and may not be intercorrelated. In contrast traits refer to consistent patterns of perception, behavior, affect, and thinking, though a certain degree of flexibility in behavior patterns still exists (but see Mischel, 1968). In other words, traits are less affected by situational changes, and as such are more stable over time. Traits can be measured directly and are often intercorrelate d for certain clusters of behavior. There is some evidence that traits and mo tives interact such that traits condition the expression of motives (see Winter, 1996; Wi nter et al., 1998). A few studies have reported that goal orientation (motives) is a mediating construct between personality traits and outcomes (Ellio tt & Church, 1997; Zweig & Webster, 2004). There is also evidence that motives are subsumed with in traits (Borkenau, 1990; Hofstee, 1994; McCrae, 1994; McCrae & Costa, 1996; Oste ndorf & Angleitner, 1994; Read, Jones, & Miller, 1990). Since the literature contends that persona lity traits have a fundamental impact on motives, although the magnitude of this inte raction remains uncertain, it is quite conceivable that personality tr aits will both interact with strain and condition the expression of strain. If this is the case, a test of the effect of strain and personality traits on delinquency, including inter action terms of strain and personality traits, should provide a meaningful analysis toward understa nding the dynamics of these relationships. That is, if the strain by tr aits interaction measure si gnificantly affects delinquency, support for the conditioning argument of persona lity traits and GST will be increased. It is also conceivable th at strain, as a proxy measur e of motive, is actually subsumed within personality traits. Such a finding would suggest that traits, rather than

PAGE 72

61 strain, are predictors of crime. If this is the case, a test of the effect of strain on delinquency while controlling for relevant pers onality traits should determine if strain affects delinquency for reasons related to pe rsonality traits. If the strain measure continues to significantly aff ect delinquency after controlling for such factors, support for GST will be increased. Furthermore, traits and strain may have reciprocal effects on one another, with each conditioning the effects of the other on de linquency/crime. Tests of this argument would require the use of longitudinal data in cluding measures of st rain and personality for at least two separate points over time. If th e effects of personality traits and strain at Time 2 are significantly related to personality traits and stra in at Time 1, and vice versa, support for a reciprocal argument would be ga ined. Since GST is based on motivations toward crime, it seems reasonable for research ers to empirically examine the influence of personality. Personality Traits Personality is a rather ambiguous concept that is defined in a myriad of ways, depending upon the theoretical paradigm and background of the researcher studying it. Some scholars acknowledge that the concept of personality refers to patterns of thoughts, feelings, and actions; other academics define pe rsonality as characteristics that make an individual’s behavi or predictable to others. The accur acy of the definition itself has not been viewed as vital to the study of its deve lopment, dimensions, or differences. Simply put, personality is the distinctive quality or character that defines individuals as themselves. Personality operates at both a conscious and unconscious level and is both dynamic and relatively stable (see Caspi & Bem, 1990; Ge & Conger, 1999; Roberts &

PAGE 73

62 DelVecchio, 2000; Soldz & Vaillant, 1999). The st ability of personality traits depends on several factors, such as genes (e.g., Mc Gue, Bacon, & Lykken, 1993), environment (e.g., McNally, Eisenberg, & Harris, 1991; Roberts, Block, & Block, 1984), internal factors (e.g., Asendorpf & Aken, 1991; Clausen, 1993; Helson, Stewart, & Ostrove, 1995; Pals, 1999; Schuerger, Zarrella, & Hotz, 1989), and the ability of the individual to adjust to or fit into the environment (see Caspi, Elde r, & Bem, 1988; Caspi & Roberts, 1999). A large body of literature has shown that personality traits are determined in part by inheritance and/or genes, which accord ing to some researchers accounts for approximately half of the explained varian ce (ranging from 0.40 to 0.80) in personality (e.g., Bock & Goode, 1996; Carey & Gold man, 1997; Eley, 1998; Gottesman & Goldsmith, 1994; Lykken, 1995; Mednick, Gabr ielli, & Hutchings, 1984; Moffitt, 1987; Plomin & Nesselrode, 1990; Rutter, 1996; s ee also, Walsh, 2000). Research has also indicated that personal ity is determined by factors other than inheritance, such as sociocultural determinants (e .g., parenting styles, attachment to others, religion, politics, education, and income), learning mechanisms, and rational choice. In one such study, Roberts and DelVecchio (2000) conducted a meta-analysis that examined rank-order consistency (i.e., whethe r groups of individuals report the same rank ordering of measures over time) of both temperament and personality traits across longitudinal studies. The purpose of their st udy was to examine stab ility and variability of normal personality over the life-course. Their an alyses, employing estimated population test-retest correlations, revealed a linear and increasing trend in stability for personality traits over the lif e-course, reaching a peak around age 50. Interestingly, the findings demonstrated a slight dip in rank-order consiste ncy around adolescence. These

PAGE 74

63 reduced correlation coefficients corroborate fi ndings from other l ongitudinal studies of the continuity of personality (correlations range from .32 to .41: Carmichael & McGue, 1994; Haan, Millsap, & Hartka, 1986; Stei n, Newcomb, & Bentler, 1986; Stevens & Truss, 1987), implying that adolescence is a phase marked by ch anges in individual behavioral characteristics. Roberts and DelVecchio (2000) also indicated that the population correlation effect sizes were more consistent for personality traits than temperament, even after cont rolling for the time span of th e longitudinal study (r ranges were .41 to .55 and .35 to .52, respectively). In addition, maladaptive (i.e., non-confor ming) personality traits have been reported to demonstrate relative stability over time (Lenzenweger, 1999). By examining 250 subjects in the Longitudi nal Study of Personality Di sorder (LSPD) across three assessment waves, Lenzenweger investigated th e stability of personality disorders (PD). The results revealed both individual differe nce stability and mean level stability, though some change occurred over time. These findings came from one of the only studies that examined the stability and change of maladap tive personality traits. As illustrated, these results suggested stability even among these types of personality traits. Temperament Temperament refers to a moderately consistent (Asendorpf, 1992; Kagan, 1989; Kochanska, Murray, & Coy, 1997; Mathe ny, 1989; McDevitt, 1986) behavioral disposition. Such dispositions are linked to one’s inherent biological functioning and environmental characteristics during ear ly childhood (see Buss & Plomin, 1975; Goldsmith, 1996; Pedlow, Sanson, Prior, & Ob erklaid, 1993; Rothba rt & Bates, 1998;

PAGE 75

64 Thomas & Chess, 1977). As an individual matures over the life-course, temperaments begin to be transformed into more cohe sive and stable pers onality traits. Empirical evidence has begun to connect temperament with the development of adult personality traits (Block, 1993; Block & Kremen, 1996; Caspi & Silva, 1995; Cohen, 1996; see also Ahadi & Rothbart 1994; Digman & Shmelyov, 1996; Martin, Wisenbaker, & Huttunen, 1994; Wachs, 1994). Yet, these studies have limitations and often have shown only modestly significant co rrelations. Such re search suggests that personality traits may depend, in part, on th e consistency of initial temperaments. Hierarchy of Personality Since multiple factors can affect personality development, personality theorists have taken numerous approaches when exam ining personality differences. Personality traits can be measured directly and are of ten intercorrelated for certain clusters of behavior. Trait theorists often describe persona lity traits according to levels ranging from very broad to more specific char acteristics: that is, (a) supe rfactors, (b) primary factors, and (c) specific behavior events (Furnham & Heaven, 1999). Superfactors describe broad cl usters or domains of persona lity traits that can be sub-divided into smaller correlated units of analysis, primary factors, and, on an even smaller scale, specific behavior events. S uperfactors are intende d to be independent elements that serve as the fundamental build ing blocks of personalities. Psychologists debate over the most appropriate number of superfactors necessary to describe personality types; some suggest a threefactor model (see PEN: Eysenck, 1977, 1992; Tellegen, 1985), while others suggest five factors (see Five-F actor Model (FFM):

PAGE 76

65 McCrae & Costa, 1990; McCrae & John, 1992; Wiggins, 1996), six factors (Hogan, 1986), or seven factors (Benet & Walker, 1992; Cloninger, Svra kic, & Przybeck, 1993). The various models of personality differ not only in their specification of the number of factors necessary to describe person ality, but also in the way that the measures are derived (i.e., biological, communi cation indicators, mood scales, and pharmacological). Despite differences in the number of factors a nd the measurement of traits, the different models of personality demonstrate a considerable amount of conceptual overlap among the overall mode ls. For example, the FFM dimension of Agreeableness and Tellegen’s Negative Emotionality map onto very similar domains, and the FFM dimension of Conscientiousness a nd Tellegen’s Constraint map onto similar domains (for a discussion, see Miller & Lynam, 2001). Regardless of which superfactor mode l is implemented, the model should describe independent personality domains and include all aspects of behavior. Contrary to the assumption of independence among “superf actors”, studies indicate that different superfactor models often contain overlapp ing constructs (Block, 1995; Church, 1994; Lilienfeld, 1999; Watson, Clark, & Harkness, 1994). However, factors describing extraversion appear to be consistently meas ured regardless of which superfactor model is employed (see Winter et al., 1998). Primary factors are sub-divisi ons within a superfactor that reflects interrelated, yet somewhat distinct, factors. For instance, th e superfactor extraversion can be described as containing several primary factors, such as impulsivity and sociability (Furnham & Heaven, 1999). From an even smaller unit of analysis, primary factors can be

PAGE 77

66 characterized as being comprised of specific beha vior events. That is an individual that acts without thinking may be la beled as impulsive. Personality Traits Affecting Deviance It has been well established that pers onality dispositions are associated with antisocial, delinquent, and crim inal behavior (e.g., Binder, 1988; Caspi et al., 1997; Caspi et al., 1994; Cleckley, 1941; Cloninger, 1987; Eysenck, 1977; Eysenck & Eysenck, 1985; Eysenck & Gudjonsson, 1989; Farrington, 1986, 1992; Gough & Peterson, 1952; Luengo, Otero, Carrillo-de-la-Pea, & Mirn, 1994; Mak, Heaven, & Rummery, 2003; Miller & Lynam, 2001; Raine, 1993; Robins, 1966; Ru tter & Giller, 1983; Schuessler & Cressey, 1950; Tennenbaum, 1977; Tremblay, Pihl, Vitaro, & Dobkin, 1994; Waldo & Dinitz, 1967; Wilson, Rojas, Haapanen, Duxbury, & Steiner, 2001; Zuckerman, 1989). Toward this end, Krueger and associates (Krueger et al., 1994) found th at low behavioral Constraint and high Negative Emotionality we re significant predicto rs of self-reported, informant reported, and officially recorded measures of delinquency (cf. Ge & Conger, 1999; Krueger, Caspi, Moffitt, Silva, & Mc Gee, 1996). In addition, impulsivity (e.g., Farrington, Loeber, & Kammen, 1990; Gerbi ng, Ahadi, & Patton, 1987; Luengo, et al., 1994; Royce & Wiehe, 1988; White et al., 1994 ), psychoticism as defined by Eysenck and Eysenck (1976) (e.g., Furnham & Thom pson, 1991), extraversion (e.g., Furnham, 1984), neuroticism (e.g., Heaven, 1996; Silv a, Martorell, & Clemente, 1986), and sensation-seeking (e.g., Newcomb & McGee, 1991; Sim & Perez, 1991; Zuckerman, 1979, 1994) have all been shown to be significan tly associated with antisocial behaviors such as conduct problems delinquency, and criminal behavior. Conversely, Agreeableness (e.g., Heaven, 1996) and Conscientiousness (e.g., Heaven, 1996) have

PAGE 78

67 been reported as inversely associated with antisocial behavior a nd delinquency. These findings have demonstrated consistency acro ss populations, as pers onality traits have been significantly associated with an tisocial and delinquent behavior in both institutionalized and non-ins titutionalized samples (e.g., Romero, Luengo, & Sobral, 2001). Personality traits have also been li nked to alcohol and drug use (e.g., Block, Block, & Keyes, 1988; Caspi et al., 1997; Ma sse & Tremblay, 1997; Wilson et al., 2001). Studies have indicated that the use of alc ohol and illicit drugs se rves as one of the cognitive motivators to reduce the effects of negative affect among adolescents (Cooper, Frone, Russell, & Mudar, 1995; Loukas, Krull, Chassin, & Carle, 2000; Newcomb et al., 1988; Stewart, Karp, Pihl, & Peterson, 1997). Consequently, scholars have proposed that individuals who possess persona lity traits that are highly affected by intense emotions may be more susceptible to alcohol and substance use. Recently, Miller and Lynam (2001) publishe d results from a meta-analysis of 59 studies examining the relationship between the four leading models of personality (i.e., the FFM model, the PEN model, Tellegen’s three-factor model, and Cloniger’s sevenfactor model) and antisocial behavior (defined as official, parent-, teacher-, and/or selfreported crime/delinquency and antisocial person ality disorder (APD) symptoms). Their findings indicated that dimensions (or sim ilar dimensions across the four personality models) of Agreeableness and Conscientious ness were moderately and significantly related to antisocial behavior Dimensions of Extraversi on and Neuroticism ranged from non-significant to weak associations with an tisocial behavior. Op enness to Experience dimensions were not significantly re lated to antisocial behavior.

PAGE 79

68 Research has shown that adolescents ch aracterized as possessing “difficult temperaments” (i.e., easily frustrated, hyperac tive, irritable) were more likely to use alcohol and drugs (Giancola & Parker, 2001; Lerner & Vi cary, 1984; Windle, 1991). For example, studies have also examined the influence of the personality trait Negative Emotionality on alcohol and drug use (Cas pi et al., 1997; Chassin, Pillow, Curran, Molina, & Barrera, 1993; Colder & Ch assin, 1997; Labouvie, Pandina, White, & Johnson, 1990; Shoal & Giancola, 2001; Tarter Blackson, Brigham, Moss, & Caprara, 1995; Wills, Sandy, Shinar, & Yaeger, 1999), an d found that Negative Emotionality was a risk factor for alcohol and drug use. So me studies, however, have found contradictory results regarding the relationship between Negative Emotionality and substance use (Clark, Parker, & Lynch, 1999; Stice & Gonzales, 1998; cf Shoal & Giancola, 2003). Research has also demonstrated significant re lationships between othe r personality traits and alcohol and drug use, such as low Constr aint (i.e., impulsivity ) (Ge & Conger, 1999; Krueger et al., 1996; McGue, Slutske, & I acono, 1999) and Positive Emotionality (Colder & Chassin, 1997; Wills et al., 1999). Relatedly, alcohol and illicit drug use ha ve been associated with antisocial behavior. A number of studi es have demonstrated that delinquency was significantly associated with substance use (e.g., Elliott, Huizinga, & Menard, 1989; Fergusson, Lynskey, & Horwood, 1994; Gillmore et al ., 1991; Osgood, Johnston, O’Malley, & Bachman, 1988). Longitudinal research has si milarly indicated that an early onset of conduct problems created a high risk of developing substance use problems (Brook, Cohen, Whiteman, & Gordon, 1992; Dobkin, Tremblay, Masse, & Vitaro, 1995; Fergusson & Lynskey, 1998; Pulkkinen, 1983; Windle, 1990). Recently, Lynam,

PAGE 80

69 Leukefeld, and Clayton (2003) reported resu lts from a study of personality, antisocial behavior (i.e., fighting, theft, truancy, vandalism), and substance use (i.e., tobacco, alcohol, and marijuana). They suggested that the personal ity profiles for substance use and antisocial behavior were similar, such that those persons who were highly antisocial and reported a large amount of substance use were also high in neuroticism and thrillseeking, but low in agreeableness, conscien tiousness, positive emotionality, and warmth. Clearly, the study of personality traits and temperaments has contributed to the research on deviance. Psychopathic Features Affecting Deviance Certain personality traits may be more maladaptive and criminogenic than others. One segment of the population that demonstr ates severe antiso cial behavior is psychopaths, or those people characterized as possessing psychopathic personality features. Historically, the term “psychopa thic” has been around since the early 1800s, but it is only within the past fifty years that the concept has acquire d a narrower meaning. It was not until the mid-1900s, following the work of Cleckley (1941, 1976, 1982), that the contemporary concept of “psychopa thic” was formed. Cleckley’s (1941) The Mask of Sanity contained the findings of his seminal work with psychopathic patients and included detailed descriptions of their psyc hological attributes. Interspersed in the clinical descriptions of his patients, Cleckley listed sixteen attributes of a psychopath, such as egocentricity, dishonest y, a lack of remorse, and superficial charm. Cleckley characterized such psychopaths as lacking “normal” emotions. Many years later, Robert Hare expanded Cleckley’s sixteen attributes of a psychopath to twenty-one (see the Psychopat hy Checklist (PCL) (Hare, 1980)), which

PAGE 81

70 was later revised to a list of twenty char acteristics (see Psycho pathy Checklist-Revised (PCL-R) (Hare, 1991)). Hare ’s operationalization of the characteris tics that define psychopaths includes such symptoms as shallow emotions, lack of empathy, deceitful and manipulative, glib and superfic ial, egocentric and grandiose, lack of remorse or guilt, impulsivity, poor behavioral controls, and thrill-seeking (Hare, 1993). His Psychopathy Checklist (PCL) and Psychopathy Checklist-R evised (PCL-R) have demonstrated considerable empirical relia bility and validity (e.g., Hare, 1991; Hart & Hare, 1989; Salekin, Rogers, & Sewell, 1996). Psychopathic personality or psychopat hy is a stable cond ition (see Blonigen, Carlson, Krueger, & Patrick, 2003 for suppor t of heritability of psychopathy) characterized by aggression, dishonesty, impulsi vity, a lack of empathy for others, and egocentricity, particularly among males (C leckley, 1941; Hare, 1991). Psychopathy is characterized by a cluster of behavioral, interpersonal, a nd affective features (Hare, 1991). According to Lyna m and Gudonis (in press): Behaviorally, the psychopath is an impulsi ve, risk-taker involved in a variety of criminal activities. Interpersonally, the psychopath has been described as grandiose, egocentric, manipulative, for ceful, and cold-hearted. Affectively, the psychopath displays shallow emotions, is unable to maintain close-relationships, and lacks empathy, anxiet y and remorse. (p. 4) It is estimated that at least 1 percent of the general population in North America can be characterized as psychopaths (H are, 1996, 1998b). Among forensic populations (i.e., those having formal contact with the criminal justice syst em), these prevalence estimates are dramatically higher: for adult males, about 10 to 30 percent (Hare, 1991);

PAGE 82

71 for adult females, roughly 15 percent (Salek in, Rogers, & Sewell, 1997; Salekin, Rogers, Ustad, & Sewell, 1998); and for adolescen ts, around 30 percen t (Brandt, Kennedy, Patrick, & Curtin, 1997; Forth, 1995; Forth, Hart, & Hare, 1990). In reference to the size of the population of psychopaths in North Am erica, Hare (1993) said the following: To give you some idea of the enormity of the problem that faces us, consider that there are at least 2 million psychopaths in North America; the citizens of New York City have as many as 100,000 psychopaths among them. And these are conservative estimates. Far from being an esoteric, isolated pr oblem that affects only a few people, psychopathy touches virtually every one of us. (p. 2) Limitations certainly range across this body of work. Research suggests that, overall, psychopathy is a conditi on that affects individuals re gardless of gender (Bolt, Hare, Vitale, & Newman, 2004; Rutherford, Cacciola, Alterman, & McKay, 1996; Salekin et al., 1997; Salekin et al., 1998; Vitale, Smith, Brinkley, & Newman, 2002), race (Brandt et al., 1997; Cooke, Kosson, & Mich ie, 2001; Kosson, Smith, & Newman, 1990; see also meta-analysis of Skeem, Edens, Ca mp, & Colwell, 2004), or ethnicity (Compton et al., 1991; Cooke, 1997; Cooke & Michie, 1999; Gonalves, 1999; Hobson & Shine, 1998; Molt, Poy, & Torrubia, 2000; Rei ss, Grubin, & Meux, 1999). It appears, however, that the magnitude and expression of certain psychopathic features varies across sociodemographic groups. A respectable amount of research exists demonstrating the relationship between psychopathy and personality traits. Several studies have examined the relationship between the Five-Factor Model of personal ity (Neuroticism (N), Extraversion (E), Openness (O), Agreeableness (A), and Conscientiousness (C)) and psychopathy (e.g.,

PAGE 83

72 Harpur, Hart, & Hare, 1994; Lynam, 2002b; Mi ller, Lynam, Widiger, & Leukefeld, 2001; Widiger & Lynam, 1998; but see Hart & Ha re, 1994; Lynam, Whiteside, & Jones, 1999). According to Lynam and Gudonis (in pre ss), psychopathic individuals may be characterized as … extremely low in Agreeableness (i.e., suspicious, deceptive, exploitive, aggressive, arrogant, and tough-minded); extremely low in Conscientiousness or Constraint (i.e., having trouble controlling his impulses and endorsing nontraditional values and standards); and tending to experience negative emotions, particularly anger and cr avings-related di stress. (p. 21-22) Psychopathy has been shown to be significantly positively associated with extraversion and impulsivity (Blackburn & Coid, 1998; Mill er et al., 2001), and to have a negative correlation with internalizing problems (Miller et al., 2001). (Internalizing problems may be measured as negative emotions, which contradicts Lynam and Gudonis’ (in press) above characterization of psychopaths. Ho wever, internalizing problems is a broad construct which may result in either positive or negative correlations depending upon the way it is measured.) Psychopathy has also de monstrated moderate positive associations with maladaptive personality features such as antisocial, borderline, histrionic, paranoid, narcissistic, and passive-aggres sive behavior, and negative as sociations with compulsive and dependent behavior (Blackburn & Coid, 1998; Miller et al., 2001). Psychopathic individuals have demonstrated a higher tendency to be motivated by desires for revenge or retaliation (Williamson, Hare, & Wong, 1987) than nonpsychopathic individuals. They also repor ted greater tendencies toward aggression (Heilbrun et al., 1998; Hemphill, Hare, & Wong, 1998; Patrick, Edens, Poythress,

PAGE 84

73 Lilienfeld, & Benning, manuscript in preparatio n; Salekin et al., 1996) and a higher use of instrumental aggression (Serin, 1991). The concept of psychopathy is of partic ular interest when examining crime and other at-risk behavior. Research has indicat ed that psychopathic adult offenders commit a disproportionately higher amount of cr ime than non-psychopath s (Blackburn & Coid, 1998; Hare & Jutai, 1983; Hare, McPhers on, & Forth, 1988; Kosson et al., 1990; Miller et al., 2001). For example, Hemphill, Ha re, and Wong (1998) reported that psychopaths were approximately four times more likel y to commit violent crime than those not identified as psychopaths. Individuals possess ing psychopathic features are also involved in other types of risk behavior. Psychopathic features are associated with higher rates of alcohol and illicit drug use (Hemphill, Ha rt, & Hare, 1994; Miller et al., 2001; Rutherford, Alterman, Cacciola, & McKay, 1997; Smith & Newman, 1990) and high risk sexual practices (Tourian et al., 1997). Adu lts possessing psychopathic features are often more disruptive in institutiona l and correctional facilities (Forth et al., 1990; Hare & McPherson, 1984; Wong, 1984). Su ch individuals have also been shown to benefit less from treatment (Ogloff, Wong, & Greenwood, 1990; Rice, Harris, & Quinsey, 1990). In addition, adults characte rized as psychopathic are more likel y to recidivate (Harris, Rice, & Cormier, 1991; Hart, Kropp, & Hare, 1988; He mphill et al., 1998; Salekin et al., 1996) and violate conditions of treatment releas e (Alterman, Rutherford, Cacciola, McKay, & Boardman, 1998; Hare, 1981; Hare, Cl ark, Grann, & Thornton, 2000; Hare & McPherson, 1984; Hart et al., 1988; Hill, R ogers, & Bickford, 1996; Ogloff et al., 1990; Rice, Harris, & Cormier, 1992; Rice et al., 1990; Serin, 1991; Serin, Peters, & Barbaree, 1990; Wong, 1984; but see Salekin, 2002).

PAGE 85

74 Psychopathy: Taxon or Dimension? Among scholars studying psychopathy, there is a debate over the nature of the construct. Some scholars consider psyc hopathy to be dimensional or continuous, whereas others believe it to be taxonomic or categorical. In general, psychopathy has been conceptualized as a global construct that is relatively uniform and continuous (see Skeem, Poythress, Edens, Lilienfeld, & Cale, 2003). This dimensional approach is reflected by the utilizat ion of total scores on psychopa thy assessment instruments. Dimensional measures of psychopathy suggest that individuals vary by degree, rather than in kind, with respect to psychopathy (H are, 1998a). To date, research addressing which approach to operationa lizing psychopathy (categorical versus dimensional) is preferable and more accurately reflects variation in the construct is lacking (but see Harris, Rice, & Quinsey, 1994). Recent studies combined with theo retical conjecture, however, have suggested a need to reconsid er the measurement of psychopathy. A few scholars (Karpman, 1941, 1948; Po rter, 1996; Mealey, 1995a, 1995b) have postulated that the concept of psychopat hy should include two variants: primary psychopathy and secondary psychopathy. On the surface, these two variants of psychopathy share many of the same characte ristics. For instance, both primary and secondary psychopaths will demonstrate antiso cial, deceptive, hostile, and irresponsible behavior (Skeem et al., 2003). Although preliminary empirical evidence should be interpreted with caution, primary and secondary psychopathy are believed to differ with regard to the etiology and motivation of these behaviors. Primary psyc hopathy is believed to be caused by genetic factors and motivated by purposeful and uncons cionable efforts to satisfy desires. In

PAGE 86

75 contrast, secondary psychopathy is believed to be caused by negative psychosocial and environmental conditions (e.g., abuse, parental rejection/neglect) (F orth & Burke, 1998; Margolin & Gordis, 2000; Marshall & Cooke, 1995; Porter, 1996; Weiler & Widom, 1996) and motivated by emotional and impulsi ve responses to negative environmental circumstances. Research examining the heterogeneity of psychopathic features has revealed findings that may further substantiate the theoretical premise that psychopathy can be described as either primary or secondar y in nature. Although the total score of measurement items is usually used to dia gnose psychopathy (e.g., for PCL-R a total score of 30 or higher diagnoses psychopathy (Har e, 1991)), studies have shown that psychopathic features cluster nicely into separate interre lated groupings of psychopathic traits. Some scholars suggest that psychopathy is best explained by two factors (see Harpur, Hakstain, & Hare, 1988, but see Hare & Neumann, 2 005 for a four-factor model of psychopathy), where Factor 1 refers to th e affective and interpersonal features (e.g., callousness, glibness, manipulativeness, sh allowness) and Factor 2 refers to the behavioral, antisocial features (e.g., aggression, impulsivity, irresponsibility). Others have recommended three factors for psychopath y (see Cooke & Michie, 2001), such that Factor 1 refers to the interper sonal styles, Factor 2 refers to the affective features, and Factor 3 refers to the impulsive and irresponsible behavior features. Studies of the subscales or factors of psychopathy (typically conducted on the two factors model of the PCL-R) have demonstrated distinct correlations with other measures. Factor 2 has been positively associated with measures of neuroticism, negative emotionality, and anxiety, while Factor 1 has been negatively associated with these

PAGE 87

76 measures (Blackburn & Coid, 1998; Frick, Li lienfeld, Ellis, Loney, & Silverthorn, 1999; Hare, 1991; Harpur, Hare, & Ha kstian, 1989; Patrick, 1994; Patr ick et al., manuscript in preparation; Verona, Patrick, & Joiner, 2001; but see Schmitt & Newman, 1999). Factor 1 has been negatively correlated with ps ychopathological traits of avoidant and dependent behavior (Blackburn & Coid, 1998). Factor 1 has been associated positively with extraversion, and negatively with personal ity traits of introversion and neuroticism (Blackburn & Coid, 1998). Factor 2 has been positively associated with anger, emotional reactivity, impulsivity, sensation-seeking, and psychopathic devi ance, and negatively associated with conscientiousness and constr aint (Blackburn & Coi d, 1998; Hare, 1991; Harpur et al., 1989; Patric k, 1994; Patrick, Bradley, & Lang, 1993; Verona et al., 2001). Moreover, studies have examined differen ces between the relationships of the two factor model of psychopathy and criminality and treatment. In a meta-analysis of psychopathy and recidivism, Hemphill et al. (1998) reported Factor 2 was a greater predictor of general recidivism, while bot h Factor 1 and Factor 2 predict violent recidivism. Further, Factor 1 has been show n to be associated with recidivism for sex offenses (Seto & Barbaree, 1999). Factor 2 has been shown to be asso ciated with alcohol and drug dependency among criminal offenders with high levels of psychopathy (Patrick et al., manuscript in preparation; Smith & Ne wman, 1990). Factor 1 has been associated with poor psychological treatment perfor mance (Hughes, Hogue, Hollin, & Champion, 1997) and disruptive behavior in treatment m eetings (Hobson, Shine, & Roberts, 2000). The aforementioned research has demonstr ated that variations exist in score configurations across psychopat hy factors and significant associ ations with these distinct factors of psychopathic features. These findi ngs may be useful in distinguishing between

PAGE 88

77 primary and secondary psychopathy. Moreover, examination of the dimensions that underlie psychopathy, in particular affective (e.g., callousness) versus behavioral (e.g., impulsivity/irresponsibility), may lead to a better understanding of the etiology of antisocial behavior, especially delinquency. Juvenile Psychopathy In an effort to better understand the e tiology and stability of severe antisocial behavior, researchers have begun to extend the construct of psychopathy downward to populations of children and adolescents. Some scholars have raised ethical concerns about extending the concept of psychopathy downward (e.g., Edens, Skeem, Cruise, & Cauffman, 2001; Quay, 1987; Seagrave & Gr isso, 2002; Steinberg, 2001; but see Frick, 2002; Hart, Watt, & Vincent, 2002; Lynam, 2002a). Due to limited prospective longitudinal research, these scholars cauti on that the implications of adolescent psychopathy are premature and may be an uncer tain predictor of lif e-course criminal propensity. They warn against hastily la beling certain adolescents as fledgling psychopaths and potentially providing poor pr ognoses of life-course psychopaths. Moreover, they argue that, unlike psychopat hy among adult populations the literature on adolescents contains limited and less reliable research pertaining to negative treatment outcomes. In particular, Seagrave and Grisso (2002) have suggested that developmental changes that occur during adoles cence may be mistaken for ps ychopathic characteristics. For these reasons, some schol ars stress that evidence th at psychopathy exists in adolescents similar to that in adults should be critically examined. Lynam and Gudonis (in press) offered tw o counterclaims to the aforementioned criticisms of the downward extension of psychopathic assessment. In response to

PAGE 89

78 concerns that researchers examining psyc hopathy in youths may actually be examining developmental changes rather than stable psychopathic features, Lynam and Gudonis referred to research that has emphasized both relative and absolute stability levels of psychopathy across adolescence (Frick, Corne ll, Barry, Bodin, & Dane, 2003; Lynam et al., in press). In response to critics’ c oncerns that examination of psychopathy among juveniles may result in the misguided applic ation of a negative la bel for youths found to possess such features, Lynam a nd Gudonis (in press) contended that it is not so much the label researchers should be concerned about as it is the mindset among some researchers the psychopathy means “untreatab le.” They emphasized that the focus of research should be on early intervention and treatment of ps ychopathy, before the development of other reinforcing negative consequences (e.g., asso ciation with delinquent peers, reduced family attachment, substance use, etc.). Such findings underscore that while psychopathy has been shown to be stable over time, it is not completely resi stant to change. The impressive volume of recent contributions to the literature on juvenile psychopathy illustrates the interest and importa nce placed on this topic across several paradigms of research. For example, the journal of Law and Human Behavior recently devoted half of one of its issues (volume 26, issue 2) to the discussion of whether or not psychopathy should be examined at a juvenile level; the journal of Behavioral Sciences and the Law recently devoted two special issues (volumes 21 and 22) to the topic of juvenile psychopathy; and the journal of Criminal Justice and Behavior devoted a special issue (volume 28, issue 4) to psychopathy and risk assessment. Although it is prudent to heed the warnings of critics, knowledge a nd understanding of juvenile psychopathy can only be acquired through scientific pursuit.

PAGE 90

79 As mentioned, numerous studies have ex amined juvenile psychopathy. Several studies have examined the reliability (For th & Burke, 1998; Frick et al., 2003; Lynam, 1997; Lynam et al., in press; Spain, Dougl as, Poythress, & Epstein, 2004; Vitacco, Rogers, & Neumann, 2003) and validity (C orrado, Vincent, Hart, & Cohen, 2004; Falkenbach, Poythress, & Heide, 2003; L ee, Vincent, Hart, & Corrado, 2003; Murrie & Cornell, 2002; O’Neill, Lidz, & Heilbr un, 2003; Ridenour, 2001; Rogers, Johansen, Chang, & Salekin, 1997; Salekin, Leistico, Neumann, DiCicco, & Duros, 2004; Vitacco et al., 2003) of youth psychopathy measures with promising results. In general, studies examining the reliability and validity of j uvenile psychopathy assessments have been more consistent and have reported stronger findings when examining the total psychopathy scores than subscal e scores of psychopathy. Studies have indicated that youths experiencing psychopa thic traits were more likely to exhibit antisocial behavior. Juve nile psychopathy has been associated with aggression (Brandt et al., 1997; Frick, O’Brien, Wootton, & McBurnett, 1994; Lilienfeld & Andrews, 1996; Lynam, 1997; Murrie, Co rnell, Kaplan, McConville, & Levy-Elkon, 2004; Rogers et al., 1997; Toupin, Mercier, Dery, Cote, & Hodgins, 1995), use of instrumental aggression (Stafford & Corne ll, 2003), and expectat ions to experience positive rewards for use of aggression (Pardi ni, Lochman, & Frick, 2003). Juveniles with psychopathic features have exhibited both violent and non-violent offending (Campbell, Porter, & Santor, 2004; Corra do et al., 2004; Forth et al ., 1990; Kosson, Cyterski, Steuerwald, Neumann, & Walker-Matthews, 2002; Lynam, 1997; Salekin et al., 2004). Juvenile psychopathy has been associated wi th an earlier age of onset for delinquency (Campbell et al., 2004; Corrado et al., 2004) and the prediction of early delinquent

PAGE 91

80 behavior (Lynam, 1997). Juvenile psychopaths are at a greater risk of become repeat offenders (Brandt et al., 1997; Catchpol e & Gretton, 2003; Corrado et al., 2004; Falkenbach et al., 2003; Forth et al., 1990; Gretton, McBride, Hare, O’Shaughnessy, & Kumka, 2001; Toupin et al., 1995). The predictive validity for non-violent recidivism, however, has been weaker among juvenile ps ychopath samples than samples of adult psychopaths (Forth et al., 1990). Psychopathic youths have demonstrated poorer treatment outcomes (e.g., shorter span of participation, slower progress th rough different phases of treatment, poorer participation, and less clinical improvement) and greater institutional infractions (Brandt et al., 1997; Campbell et al., 2004; Forth et al., 1990; Hicks, Rogers, & Cashel, 2000; Murrie et al., 2004; O’Neill et al., 2003; Roge rs et al., 1997; Spain et al., 2004; Stafford & Cornell, 2003). Some studies have re ported that psychopathic youths experience problems related to alcohol and substance us e/abuse (e.g., earlier age of onset, variation in substances used) (Campbell et al., 2004; Corrado et al., 2004; Mailloux, Forth, & Kroner, 1997; Murrie & Corne ll, 2002; Toupin et al., 1995; but see Brandt et al., 1997; O’Neill et al., 2003). Research has produced mixed results on the association between psychopathic traits and other psychosocial problems (e .g., child maltreatment, family influences, parental attachment, peer rejection, school pr oblems) (see Campbell et al., 2004; Corrado et al., 2004; Forth & Tobin, 1995; Kosson et al ., 2002; O’Neill et al ., 2003; Piatigorsky & Hinshaw, 2004; Wooton, Frick, Shelton, & Silv erhorn, 1997). Nevertheless, research has suggested that the impact of certain psychosocial problems, like family influences, depends on the presence of callous-unemotiona l (CU) psychopathic features in youths.

PAGE 92

81 When these features were present, youths we re at greater risk fo r antisocial behavior, even in the absence of adverse family factors. When CU features were absent, family factors had a greater influence in the devel opment of antisocial behavior. Psychopathic features have been shown to be adequate predictors of antisocial behavior, when controlling for the influence of measures such as intelligence, social status, prior delinquency, impulsivity and other conduct/d isruptive behaviors (Frick et al, 2003; Gretton, Hare, & Catchpole, 2004; Lynam, 1997; Murrie et al., 2004). Studies of psychopathy among youths have also revealed a relationship between psychopathy and personality traits. Resear ch has found a strong negative relationship between juvenile psychopathy and traits of Agreeableness and Cons cientiousness (similar to Constraint) (Lynam, 2002b; Lynam et al., in press; Salekin, Leis tico, Trobst, Schrum, & Lochman, in press). These studies have al so reported significant associations between juvenile psychopathy and Neuroticism (sim ilar to Negative Emotionality), but the direction of this relationship was uncertain. Two of the stud ies (Lynam et al., in press; Salekin et al., in press) reported a modera te positive relationship between psychopathy and Neuroticism, while the other (Lynam 2002b) found a negative relationship between psychopathy and Neuroticism. In addition to experiencing psychopathic symptoms, it has become established that many adolescents are experiencing othe r psychological problems. Research has indicated a significant positive associa tion between juvenile psychopathy and externalizing behaviors (i.e., disruptive behaviors such as oppositional defiant disorder (ODD), conduct disorder (CD), antisocial pe rsonality disorder (APD), and substance abuse/dependency) (Brandt et al., 1997; Cam pbell et al., 2004; Frick, 2000; Frick, Bodin,

PAGE 93

82 Barry, 2000; Hume, Kennedy, Patrick, & Partyk a, 1996; Lynam, 1997; Myers, Burket, & Harris, 1995; Piatigorsky & Hinshaw, 2004). So me of these studies indicated a weak, but significant, positive association between juvenile psychopathy and internalizing behaviors (e.g., anxiety, depression, somatic sy mptoms, withdrawal) (B randt et al., 1997; Lynam, 1997; but see Campbell et al., 2004) Lynam (1997), in contrast, observed negative correlations between internalizing behaviors and psychopathy after controlling for general psychopathology. Overall, youths possessing psychopathic traits appeared to have a propensity for externalizing probl ems, but were relatively unaffected by internalizing problems. Similar to adult psychopathy assess ments, scholars have examined the consistency and validity of f actor structures in juvenile assessments of psychopathy. Among juvenile psychopathy measures, two (e .g., Frick et al., 1994) three (e.g., Vitacco et al., 2003), and four (e.g., Forth, Kosson, & Hare, 2003) factor structures have been reported. Findings regarding reliability and va lidity have been less consistent and weaker than those for the total score, but nonetheless promising. Th e research to date, however, has indicated some general weakness in juvenile psychopathy research (see Lynam & Gudonis, in press). Disagreement remain s over which factor structure optimally describes psychopathy (this remains true for adult psychopathy as well). Overall, the factor subscales were less reliable than usi ng the total score. Furthermore, convergence across the factor subscales of assessments has been rather weak. More research on juvenile psychopathy factor structure is ce rtainly needed to address these issues. An examination of psychopathy factor dimensions among juveniles has also revealed findings similar to those in adu lt studies. Utilizing data obtained from two

PAGE 94

83 cohorts from the Pittsburgh Youth Study (L oeber, Farrington, Stouthamer-Loeber, & Kammen, 1998), Lynam and colleagues (in pres s) studied the rela tionship between the two-factor model of adolescen t psychopathy and the Five-Fact or Model of personality. They found significant negative relationships between Factor 1 and Agreeableness, Neuroticism, and a positive association betw een Factor 1 and Openness. Factor 2 was positively related to Neuroticism and Extr aversion, and negatively related to Agreeableness, Conscientiousness, and Openness. The affective/interpersonal factor (Factor 1) has been associated with recidi vism (Falkenbach et al., 2003), prior violent convictions (Corrado et al ., 2004), and program non-compliance (Falkenbach et al., 2003). The behavioral factor (F actor 2) has been associated with an earlier age of onset and more severe offending (Corrado et al., 2004 ), recidivism (Falke nbach et al., 2003), school misconduct (Corrado et al., 2004) a nd program non-compliance (Falkenbach et al., 2003). The Tautology of Personality Traits and Psychopathic Features and Crime Personality theories assume that crimin ality is a “symptom” of a larger problem within the individual (Akers, 1997, p. 53). According to Akers, personality theories assume that “delinquents and criminals ha ve abnormal, inadequa te, or specifically criminal personalities or personality traits that differentiate them from law-abiding people.” (p. 53). From a criminological pe rspective, such a generalization about personality and crime suggests a tautologica l issue conceptuall y. Assuming that delinquent individuals are by definition funda mentally flawed, such that their very identities are deviant, how can science sepa rate the crime from the criminal? Based on

PAGE 95

84 the conceptualization such an act would be virtually impossible, particularly when examining severely maladaptive personality types. It is not the intention of this paper to take a stand e ither for or against arguments about the conceptual tautology of personality theory and crime. Rather, the point of presenting this problem is to acknowledge the existence of this i ssue when interpreting criminological research that relies upon personality theory and measurement. The conceptual tautology of personality within an etiological context for crime remains an issue for the criminological discipline to reso lve. Future efforts should be made to continue pursuing such a resolution. In addition to conceptual tautology, previ ous research on pers onality traits and crime/delinquency and psychopathy and crime/ delinquency has suffered from issues of empirical tautologies. As discussed in Chap ter 2, an empirical tautology occurs when two independent measures are found to be so hi ghly correlated with each other, that they are essentially measuring the same concept. In the case of personality traits and psychopathy, this has also been referred to as “predictor-c riterion overlap” (see Caspi et al., 1994). Some researchers (see Gottfreds on & Hirschi, 1990) have claimed that previous studies of personalit y traits and crime have not b een independently measured. For example, both the California Personali ty Inventory (CPI) and the Minnesota Multiphasic Personality Inventory (MMPI) contai n items referring to criminal activities. Yet, these scales and subscales are utili zed to predict criminal propensity. This becomes a point of contention es pecially among criminologists. When discussing the applicability of personality trai ts, particularly psychopathy, to the study of crime, many criminologists would agree w ith Harris, Skilling, and Rice (2001, p. 199)

PAGE 96

85 when they stated, “We had previously suspected that psychopathy was merely a euphemism for a lengthy history of officially recorded criminal conduct.” However, after many years of research on psychopathy, Ha rris and his associates have become convinced that psychopathic fe atures are not concepts that are subsumed under criminal deviance, but a separate phenom enon (Harris et al., 2001). Th is change of heart was the product of several studies wher e the researchers controlled fo r predictors of criminal history and other high-risk pred ictors of criminality, such as alcohol abuse, prior to introducing measures of psychopa thy into their analyses (e.g., Harris et al., 1994; Rice & Harris, 1995). Their findings indicated that psychopathy served as a unique predictor of criminality. However, their findings stil l do not resolve the problem of tautology. Despite the claims of some scholars th at psychopathy can be studied without committing tautological errors (Levenson, Kiehl, & Fitzpatrick, 1995; Lilienfeld & Andrews, 1996), one might question whether or not it is even possible to have noncriminal psychopaths. For example, Hare st ates that “…for those who are [psychopaths], crime is less the result of adverse social c onditions than of a character structure that operates with no reference to the rules and regulations of society.” (Hare, 1993, p. 85; text in brackets not original). Ha re continues to say the following: In many respects it is difficult to see how any psychopaths—with their lack of internal controls, their unconventional attitudes about ethics and morality, their callous, remorseless, and egocentric view of the world, and so forth—could manage to avoid coming into conflict with society at some point in their lives. (Hare, 1993, p. 86)

PAGE 97

86 Yet, based on findings from forensic and nonforensic studies, Hare also states that “…not all psychopaths are criminals…” (p. 86). Schneider (as cited in Cooke & Michie, 2001, p. 185) has also argued that non-crimin al psychopaths are well-represented in society. Examples of “non-criminal” psyc hopaths include individuals such as highly successful business and corporate leaders, li ke stock brokers, as well as unethical and corrupt lawyers, doctors, polit icians, and other white-colla r professionals (see Hare, 1993, 1996, 1998b; cf. Babiak, 1995, 2000). Psychopath s may thrive in ultra competitive and chaotic corporate and business envir onments, though there is no real empirical evidence to support such a claim. While some may take offense to the claim that these examples of “non-criminal” psychopaths are considered non-criminal, it s hould be noted that this is a loose interpretation of the term non-criminal. Certainly, one may argue that these morally weakened white-collar professionals are cr iminal depending upon one’s definition of criminal. One could also argue that accordi ng to other definitions of “criminal,” one would be hard pressed to find individuals that did not violate some moral or legal code. After all, how many people exceed the speed li mit or choose to keep the change left in a pay phone? For the purposes of this paper, “c riminal” will be most often be defined as the violation of the law, as defined in state statutes. Researchers have attempted to recrui t and study non-forensic psychopaths, selected from the general community populatio n, with unsuccessful results (e.g., Belmore & Quinsey, 1994; Lalumiere & Quinsey, 1996; Widom, 1977; Widom & Newman, 1985)—most of the subjects had histories of crim inal justice contact. In an effort to

PAGE 98

87 examine psychopathic features among non-o ffending populations, perhaps studies that utilize child and adolescent populations may provide such an opportunity. Studies of personality traits and psychopa thic features must be sensitive to the issue of empirical tautology. It is imperative that m easures of personality and psychopathy exclude items that conceptually ove rlap with criminality (see Lynam, 1997). Indeed, with respect to psychopathy, research ers have begun to develop new measures of psychopathy that do not include any explicitly antisocial or criminological items (see Levenson, Kiehl, & Fitzpatrick, 1995; Lilienf eld & Andrews, 1996). Moreover, every effort should be made to control for predictors of criminal history prior to examining the effects of personality traits and psychopathic features. Researchers who are aware of these methodological and measur ement issues and employ safeguards against empirical violations will benefit from the scientific integrity of their findings. The literature examining the influence of personality traits and temperaments on delinquency is longstanding. Although the body of research on psychopathy as it applies to the study of childhood and adolescent deli nquency is fairly new, it is growing at a respectable rate. The present study attempts to take advantage of th e increasing interest in employing psychopathy and personality traits as relevant explanat ions of deviance and criminality. However, it is important to examine the influence of personality on delinquent behavior within a theoretical framework. As Akers stated (1997, p. 1), “An effective theory helps us to make sense of facts that we already know and can be tested agains t new facts.” It is well-established in the GST literature that strain leads to cr ime/delinquency (e.g., Agnew & Brezina, 1997; Agnew & White, 1992; Aseltine et al., 2000; Bao, Haas, & Pi, 2004; Benda & Corwyn,

PAGE 99

88 2002; Brezina, 1999; Broidy, 2001; Hoffma nn, 2002; Hoffmann & Cerbone, 1999; Hoffmann & Miller, 1998; Hoffmann & Su, 1997; Mazerolle, 1998; Mazerolle et al., 2000, 2003; Mazerolle & Maahs, 2000; Maze rolle & Piquero, 1997, 1998; Paternoster & Mazerolle, 1994; Piquero & Sealock, 2000, 2004; Robbers, 2004; Sharp et al., 2005; Sigfusdottir et al., 2004; Wallace et al., 2005) Research has also consistently demonstrated a link between personality and an tisocial behavior (e.g., Caspi et al., 1997; Caspi et al., 1994; Cloninger, 1987; Eysenc k & Gudjonsson, 1989; Luengo et al., 1994; Mak et al., 2003; Miller & L ynam, 2001; Raine, 1993; Trembl ay et al., 1994; Wilson et al., 2001; Zuckerman, 1989) and psychopathy and an tisocial behavior (e .g., Brandt et al., 1997; Campbell et al., 2004; Catchpole & Gr etton, 2003; Corrado et al., 2004; Forth et al., 1990; Gretton et al., 2001; Kosson et al., 2002; Lynam, 1997; Salekin et al., 2004; Toupin et al., 1995). While ther e is empirical consistency rega rding the direct effects of strain on delinquency, little is really known about factors th at condition this relationship, particularly personality. Examination of personality within a GST framework may provide important clues about individual differences in the management of strainful events and conditions. In the next chapter, a detailed descri ption of Agnew, Brezina, Wright and Cullen’s (2002) test of general strain theory and personality is presented. This discussion is followed by a description of the theore tical premise of the present study, and explanation of the conceptual models to be examined.

PAGE 100

89 Chapter 4 General Strain Theory and Personali ty: Another Theoretical Elaboration As mentioned in previous chapters, A gnew and associates (2002) have recently advocated for another theoretica l elaboration of general strain theory: the examination of the effects of personality traits as conditioning factors on the strain-delinquency relationship. Part of the impetus for advanc ing this extension of GST was derived from a previously published (Agnew, 1997) theoretical discussion of GST from a developmental perspective, in which Agnew emphasizes the importance of personality traits in GST within the context of GST’s ability to explai n the stability and change in crime over the life-course. Agnew (1997) has argued that GST may play a significant role in the explanation of developmental trajectories developmental pathways that track important transitions occurring over the life-course (see Sampson & Laub, 1993, 1997), of crime. According to Agnew, GST may offer a supplemental expl anation of the stabili ty and change in crime over the life-course, specifically addr essing how criminal behavior can be characterized as “life-course-persistent”4 in some individual s and “adolescence-limited”5 in others (Moffitt, 1993). Within this cont ext, GST postulates that certain personality traits (e.g., impulsivity, hyperac tivity, difficult temperament, etc.) increase the likelihood 4 Life-course-persistent offenders are characterized as having an early age of onset, exhibiting extensive criminal behavior throughout adolescence, and continui ng such behavior into adulthood (Moffitt, 1993). 5 Adolescence-limited offenders are characterized as ha ving a later age of onset and a shorter duration of criminal activity, usually terminating with successf ul adjustment into adulthood (Moffitt, 1993).

PAGE 101

90 that individuals will e xperience strain, interpret strainful situations as aversive, and cope with these situations thro ugh criminal behavior (Agnew, 1997, p. 107). Agnew suggests that an examination of the development and stability of these pe rsonality traits as conditioning factors for strain may help to explain differences between “life-coursepersistent” trajectories and “adolescence-lim ited” trajectories. Agnew and associates (Agnew et al., 2002) have since examined the role of personality traits in GST in a more general context. Although the emphasis on pe rsonality traits has b een removed from the context of the life-course perspective under this genera l context, the theoretical propositions linking personality to the strain-delinquency rela tionship remain the same. GST and the Conditioning Effects of Personality and Psychopathic Features As the preceding chapter illustrates, a relationship between personality traits and motives seems tenable. Strain as operati onalized by GST is essentially motive-derived from both subjectively and objectively define d negative relationships. Given the link between motives and personality, it seems plau sible that Agnew et al. (2002) would have grounds for a test of GST examining the modera ting role of certain personality traits on strain. In fact, Agnew et al. recognize the negl ect of personality traits as an oversight, not only for GST, but for the criminological field in general. They even go as far as stating that “…the impact of such [personality] traits may be far more pervasive than that of the conditioning variables typical ly examined…” in the GST research, affecting how individuals (a) emotionally re spond to strain, (b) develop non -deviant coping strategies, and (c) develop deviant coping st rategies, particularly with re spect to perceptions of the costs of illegitimate responses and deviant dispositions (Agnew et al., 2002, p. 45). This argument serves as the impetus for thei r study of GST and personality traits.

PAGE 102

91 Studies have shown that pers onality traits can affect the interpretation of stress or strainful conditions (e.g., Eysenck, 1989). For example, Costa, Somerfield, and McCrae (1996) suggested that the pers onality traits (i.e., Neurotic ism, Extraversion, Openness, Conscientiousness, and Agreeableness) influe nce the ways that individuals cope with stressful conditions. For example, they found that individuals high in Neuroticism cope with stress in a more emotional manner than other personality types, such as becoming irritable or acting childish. Those high in Extraversion cope with stress by making attempts to minimize the situation by joking ab out it or discussing it with others. Those high in Openness cope by attempting to disc over creative and alte rnative methods for handling the stress. Those high in Conscien tiousness cope with stress by focusing on the task and meditating or praying for guidance. Those high in Agreeableness cope with stress via acquiescence. Other researchers ha ve demonstrated similar results comparing coping mechanisms and personality traits (e.g., Brebner, 2001; Uehara, Sakado, Sakado, Sato, & Soomeya, 1999). In another study of personality and stre ss, Wofford, Daly, and Juban (1999) found that cognitive-affective structures that are asso ciated with certain personality traits affect responses to school stress a nd physiological strain. They examined a construct of cognitive-affective stress propensity (CASP) co mposed of six personality traits (Wofford et al., 1999, p. 44-46): negative affectivity (i.e., introspectiveness), self-esteem, pessimistic attribution style, locus of c ontrol, cognitive-affective connectivity, and psychological magnification. Their findings sugg est that personality traits, particularly those correlated with expressi ons of negative affect, signifi cantly lead to stress and physiological manifestations of stress.

PAGE 103

92 Personality traits have also been linked with differences in affect. For example, Brebner (1998; cf. Brebner, 2001) examined the effect of the Big Four personality traits (i.e., Happy, Labile, Stable, and Unhappy) on positive (affection, contentment, joy, and pride) and negative emotions (anger, fear, guilt, and sadne ss). He found that Labile individuals experienced high le vels of both positive and negative emotions, while Stable individuals experienced low levels of both positive and negative emotions. Happy individuals experienced high levels of pos itive emotions, but low levels of negative emotions; and Unhappy individuals experienced the opposite. Studies have also reported a significant association between psychol ogical distress (i.e., depression, anxiety, somatization, hostility) and Tellegen’s ( 1985) Negative Emotionality domain of personality (i.e., anxiety, anger, rebelli ousness/argumentativeness) (Ge & Conger, 1999; Krueger et al., 1996). The First Test of GST and Personality Utilizing a sample of nationally representative data obtained from the second wave (in 1981) of the National Survey of Ch ildren (NSC), Agnew et al. (2002) presented a cross-sectional study examining the effects of strain, social control, social learning, and personality traits on delinquency. The samp le contained 1,423 youths, both male and female, between the ages of 12 and 16 years old. Strain was measured as family strain (e.g., family life tense/stressful not cooperative, not organi zed), conflict with parents (e.g., arguments, yelling), parental perception of loss of control, poor peer relations (e.g., picked on by peers), school ha tred, neighborhood strain, and a composite index of strain. Social control was measured as attachment to parents, firm parental discipline, school commitment, school attachment, goals for college, amount of time working on homework

PAGE 104

93 each day, and conscience (i.e., shame fo r doing something wrong). Differential association/social learning wa s measured as parental percep tion of troublesome peers. Personality traits were measured as an index of Negative Emotionality (i.e., anxiety, anger, rebelliousness/argumentativeness) a nd low Constraint (i.e., impulsivity) (see Tellegen, 1985). Delinquency was measured as an index of five items indicating selfreported assault, theft, vandalism, skipping school, and drinking/drunkenness. A measure controlling for prior (from wave one) aggression and vandalism was also included in the analyses. Agnew et al. (2002) conducte d a regression analysis examining the effects of the separate measures of strain, social cont rol, differential association, and Negative Emotionality/low Constraint on delinquenc y, controlling for prior aggression and vandalism and other sociodemographic variable s. The results indicated that several measures of strain (family strain, parental loss of c ontrol, school hatred, and neighborhood strain) were signifi cantly, positively related to delinquency. Social control as a measure of school attachment was signifi cantly, negatively re lated to delinquency. Differential association as a measure of tr oublesome friends was si gnificantly, positively related to delinquency. Negative Emotionali ty/low Constraint wa s also significantly related to delinquency, such that youths high in Negative Emotionality/low Constraint were more likely to self-report acts of delinquency. Next, the authors examined the role of Negative Emotionality/low Constraint as a moderating variable for strain. They created a composite inde x of strain using only those measures of strain that si gnificantly affected delinquency. Then, they regressed these measures of strain, social c ontrol, differential association, and Negative Emotionality/low

PAGE 105

94 Constraint on delinquency, controlling for prior delinquency and other sociodemographic variables, without and with th e inclusion of an interacti on term of strain X Negative Emotionality/low Constraint. The findings fo r the composite strain model excluding the interaction term were consistent with those of the model including separate measures of strain. The composite index of strain was si gnificantly, positively re lated to delinquency. The findings from the composite strain mode l including the strain-trait interaction term indicated that Negative Emoti onality/low Constraint significan tly conditions the effect of strain on delinquency. Strain is more lik ely to lead to delinquency among youths reporting high Negative Emotionality and lo w Constraint personality traits. The magnitude of the relationships between th e independent and control variables and delinquency were not affected by the introduction of the inte raction term, thus suggesting that the strain-delinquency relationship is not spurious with respect to certain personality traits. However, the introduction of the in teraction term did not improve much upon the explained variance of the model; the adjusted R2 increased slightly from 0.19 to 0.20 once the interaction te rm was included. The study conducted by Agnew et al. ( 2002) provides a significant, though empirically limited by the cross-sectional a pproach and limited measurement of strain and personality, contribution to the literature on GST. Stra in appears to be conditioned by personality traits/features, as the psychological literature has previously indicated. Agnew has broached the proposition that tra its may condition the effect of strain on delinquency on numerous occasions (see Agnew, 1992, 2001; Agnew et al., 2002). Current research provides pa rtial support for his contenti ons, but there is need for replication before the relative importance of personality traits within a GST context can

PAGE 106

95 be determined. The present study offers to eith er strengthen or refute the initial findings of Agnew et al. regarding the role of traits in GST by provi ding a partial replication and extension of their study using altern ative measures of personality. The Proposed Study In an article presenting results from a meta-analysis of personality models and antisocial behavior, Miller and Lynam (2001) discussed how personality can affect the development of antisocial behavior/crime. Th ey suggested that the etiology of crime may not depend solely on personality traits, despit e evidence indicating th at personality is a relatively stable inherent quality, but rather, “t he presence of a thir d variable” (p. 781). Miller and Lynam advocate for the examina tion of intervening mechanisms between personality and crime. Specifically, they suggest examining how personality traits may influence an individual’s environmen t and decision-making processes. Based on research from Caspi and Be m (1990) on personality-environment transactions, Miller and Lyna m (2001)suggest that a person’ s personality may influence how he/she interprets and responds to his/ her surroundings (i.e., r eactive transactions), how others react toward and treat him/her (i .e., evocative transactions), and which types of social environments he/she selects (i.e., pr oactive transactions). Reactive transactions refer to the way in which an individual re sponds to situations and circumstances. For example, personality traits may influen ce the means by which individuals respond to situations. A person characterized as having aggressive personality traits will be most likely to react to situations in an aggressi ve manner and believe that aggressive coping strategies will provide the most successful outcomes. Evocativ e transactions refer to the responses that one evokes from others. For example, parents may respond to children

PAGE 107

96 with difficult personalities or temperaments by using harsh and erratic discipline and reducing their interaction with the child as he gets older. Proactive transactions refer to an individual’s selection of so cial environments that are in -line with his/her personality traits. For example, people tend to choose si milar others as friends. These reactive, evocative, and proactive tran sactions describe distal wa ys in which personality may influence crime (Miller & Lynam, 2001). At a proximal level, personality may also influence immediate decision-making (Miller & Lynam, 2001). For example, individuals low in Constraint may be less likely to base decisions upon information-gathering techniques (Patterson & Newman, 1993) These distal and proximal interactions between personality and the environment may describe how personality influences the strain-delinquency relationship. As previously discussed in Chapter 2, Agnew ha s described several conditioning factors. Two of these factors relate to low social control (i.e., evoca tive transactions) and delinquent peer associations (i .e., proactive transac tions). Further, GST postulates that these conditioning factors may di rectly affect strain and in fluence the selection and/or availability of legitimate coping mechan isms (Agnew, 1992). If personality can influence the how individual’ s respond to their surroundi ngs, how others respond to them, which social networks they establish, and what they perceive as viable coping mechanisms available to them, then Agnew et al. (2002) may be justified in assuming that GST will certainly benefit from examina tion of the moderating and mediating effects of personality traits on strain and delinquency. In their analysis, Agnew et al. (2002) c hose to examine the moderating effects of two domains of Tellegen’s (1985) mood-ba sed personality trait model: Negative

PAGE 108

97 Emotionality and Constraint. Tellegen ( 1985) developed a three-factor model of personality: Positive Emotionality (PEM), Nega tive Emotionality (NEM), and Constraint (CON). PEM refers to the procli vity for individuals to socially interact with others in a positive manner. NEM refers to the inclinati on for individuals to express and experience negative emotions such as ange r, anxiety, and fear, particular ly under stressful situations. As such, NEM typically include s items comprising an aggressi on subscale as part of its assessment (Miller & Lynam, 2001, p. 779). CON refers to an individuals’ ability to control emotions, impulsivity, and rash deci sions. According to Agnew and colleagues (2002), both high NEM and low CON have been associated with delinquency; NEM, in particular, is linked with aggression. In addition to NEM and low CON, other ma ladaptive personality traits have been linked with aggression and deli nquent behavior. In particul ar, psychopathic personality traits have been linked with aggression and antisocial behavior. A respectable amount of research demonstrates the relationship be tween psychopathy and normative personality traits and the ability of pe rsonality assessments to m easure psychopathic features (especially the Five-Factor Model [FFM]) (e.g., Harpur et al., 1994; Lynam, 2002b; Miller et al., 2001; Widiger & Lynam, 1998; but see Hart & Hare 1994; Lynam et al., 1999). Although most of the studies examini ng the association between psychopathy and personality have not relied upon Tellegen’s three-factor model of personality, scholars have demonstrated that in terms of traits there exists substantial trait agreement across the most widely used personality models [i.e., FFM (McCrae & Costa, 1990; McCrae & John, 1992; Wiggins, 1996), Eysenck’s thr ee-factor model (PEN: Eysenck, 1977, 1992), and Tellegen’s three-factor model (1985)] (for a discussion see Miller & Lynam, 2001).

PAGE 109

98 As previously discussed in Chapter 3, Lynam and Gudonis (in pr ess) have stated that an individual possessing psyc hopathic traits tends to be … extremely low in Agreeableness (i.e., suspicious, deceptive, exploitive, aggressive, arrogant, and tough-minded); extremely low in Conscientiousness or Constraint (i.e., having trouble controlling his impulses and endorsing nontraditional values and standards); and tending to experience negative emotions, particularly anger and crav ings-related dist ress. (pp. 21-22) Psychopathic individuals demonstrate a higher tendency to be motivated by desires for revenge or retaliation (Williamson et al., 1987 ) than non-psychopathic individuals. They also report greater tendencies toward aggr ession (Heilbrun et al., 1998; Hemphill et al., 1998; Patrick et al., manuscript in preparation; Salekin et al., 1996) and a higher use of instrumental aggression (Serin, 1991). Psychopa thic features have also been linked to criminal behavior (Blackburn & Coid, 1998; Ha re & Jutai, 1983; Ha re et al., 1988; Harris et al., 1991; Hart et al., 1988; Hemphill et al., 1998; Kosson et al., 1990; Miller et al., 2001; Salekin et al., 1996) and alcohol and/ or illicit drug use (Hemphill et al., 1994; Miller et al., 2001; Rutherford et al., 1997; Smith & Newman, 1990). Among juveniles, psychopathy has been asso ciated with aggression (Brandt et al., 1997; Frick et al., 1994; Lilienf eld & Andrews, 1996; Lynam, 1997; Murrie et al., 2004; Rogers et al., 1997; Toupin et al., 1995), us e of instrumental aggression (Stafford & Cornell, 2003), and expectations to experience positive rewards for use of aggression (Pardini et al., 2003). Juvenile psychopathic features have been linked to both delinquent behavior (Campbell et al., 2004; Corrado et al., 2004; Fort h et al., 1990; Kosson et al., 2002; Lynam, 1997; Salekin et al., 2004) and pr oblems related to alcohol and substance

PAGE 110

99 use (e.g., earlier age of onset, variation in substances used) (Campbell et al., 2004; Corrado et al., 2004; Mailloux et al., 1997; Murrie & Cornell, 2002; Toupin et al., 1995; but see Brandt et al., 1997; O’Neill et al., 2003). In addition to experiencing psychopa thic symptoms, many adolescents demonstrate other maladaptive personality char acteristics. Historically, personality and psychopathology (i.e., abnormal personality or me ntal disorders) have been treated as empirically distinct. Some scholars have suggested, however, that psychopathology is linked to and maps onto broad personality tra it domains (e.g., Krueger, 2002; Krueger et al., 1994, 1996, 2002; Krueger, McGue, & Ia cono, 2001; Krueger & Tackett, 2003; Livesley, Schroeder, Jackson, & Jang, 1994; Watson et al., 1994). Personality models contain traits that are rele vant to a broad range of in ternalizing and externalizing psychopathological behaviors. For example, in Tellegen’s (1985) three-factor model, the CON domain has been negatively correlated with externalizing be haviors (e.g., Blonigen et al., 2003; Krueger et al., 2001), and the NE M domain has been positively correlated with internalizing behavior (e.g., Krueger et al., 2001). Moreover, in relation to psychopathy, youths possessing psychopathic tr aits appear to have a propensity for externalizing problems (Brandt et al., 1997; Campbell et al., 2004; Fr ick, 2000; Frick et al., 2000; Hume et al., 1996; Lynam, 1997; My ers et al., 1995; Piatigorsky & Hinshaw, 2004), but are relatively unaffected by inte rnalizing problems (Brandt et al., 1997; Lynam, 1997; but see Campbell et al., 2004). Not all youths experiencing strain will commit delinquent acts (including illicit substance use). According to GST (Agnew, 1992), the appropriate coping mechanisms and conditioning factors must be present for individuals to become deviant. Similarly,

PAGE 111

100 not all individuals possessing psychopathic fe atures (or even satisfying the threshold for what may be misleading labeled psychopathy—s ee discussion of primary and secondary psychopathy in Chapter 3) and/or psychopatholog ical characteristics will become deviant. It is the premise of this study, however, that youth who are high in maladaptive personality characteristics (psychopathi c features and general psychopathological characteristics) will be more likely to respond to strainful conditions with delinquency. Youths high in maladaptive personality tr aits will be more likely to view negative relationships with others as aversive and experience intense emotional reactions to these circumstances. They will also be more li kely to respond to these situations through aggression and antisocial means. Similar to Agnew et al.’s (2002) use of NEM and low CON, the present study examines the infl uence of psychopathic externalizing, and internalizing personality characteristics th at have demonstrated a tendency to be behaviorally reactive and weak in pros ocial coping mechanisms under stressful conditions and situations of criminal opportunity. GST, Personality, Delinquency, and Drug Use Problems Due to the relatively small sample size be ing used in this st udy (see discussion of the sample in Chapter 5), the study presente d must be parsimonious, and is therefore limited in the number of measures that can be examined. Data reduc tion techniques, such as factor analysis, are used to maintain model parsimony. Multivariate structural equation models (SEM) of the hypothesized re lationships between measures of strain, social control, differential association, pe rsonality/psychopathic traits, and delinquency and drug use are tested.

PAGE 112

101 Consistent with Agnew’s (1992, 2001, 2002) insistence that studies of GST include measures of social control and social learning, the present study includes measures of social contro l (parental attachment, pare ntal firmness, and school attachment) and differential association/social learning (delinquent pe er associations). While these measures are considered to be indi cators of rival theori es, low social control and low differential association are viewed as correlated with high leve ls of strain. This decision is predicated on Agnew’s (2001) specification of criminogenic strainful conditions. That is, strain accompanied by low social control and high differential association should be more crime-inducive. The literature on social control has cons istently indicated that poor parental relationships, particularly ma ltreatment, increase the risk s that a youth will become involved in delinquency and substance use (e.g., Dembo et al., 1990; Dembo et al., 1992a; Dembo, Williams, Werner, Schmeidl er, & Brown, 1992b; Ireland & Widom, 1994; Kakar, 1996; Lemmon, 1999; Smith & Thornberry, 1995; Widom, 1991; cf. Ireland, Smith, & Thornberry, 2002). Moreove r, harsh parenting styles (Smith, & Myron-Wilson, 1998; Stormshak, Bierman, McMahon, & Lengua, 2000) and ineffective parenting (Berg-Nielsen, Vikan, & Dahl 2002; McCoy, Frick, Loney, & Ellis, 1999; Wootton et al., 1997) have been linked to higher risks for aggression and antisocial behavior. Research has also suggested that the presence of deviant peers fosters the development of both delinquency and substa nce use (Dishion, Patterson, Stoolmiller, & Skinner, 1991; Elliott et al., 1989; Fergusson & Horwood, 1996; Fergusson, Woodward,

PAGE 113

102 & Horwood, 1999; Haynie, 2001; Kandal, 1973; Moss, Lynch, & Hardie, 2003; Piquero, Gover, MacDonald, & Piquero, 2005; Simons Wu, Conger, & Lorenz, 1994; Warr, 2002). Research on adolescent psychopathic traits has demonstrated a significant association between psychopathy and aggression (Brandt et al., 1997; Frick et al., 1994; Lilienfeld & Andrews, 1996; Lynam, 1997; Mu rrie et al., 2004; Pardini et al., 2003; Rogers et al., 1997; Stafford & Cornell, 2003; Toupin et al., 1995) and delinquency (Campbell et al., 2004; Corrado et al., 2004; Forth et al., 1990; Kosson et al., 2002; Lynam, 1997; Salekin et al ., 2004). Juvenile psychopathy has also been linked to substance use (Campbell et al., 2004; Corrado et al., 2004; Mailloux et al., 1997; Murrie & Cornell, 2002; Toupin et al., 1995), though not as consistently (see Brandt et al., 1997; O’Neill et al., 2003). Based on these studies, psychopathic individuals, those characterized by impulsivity, thrill-seeking, hostility, low self-control, and low empathy for others, are at risk for involvement in delinquency and substance abuse. In the previous chapter, it was reported that psychopathy has been studied in two ways: (a) as a uniform construc t (i.e., the total score) and (b ) as specific variants of the construct. According to Skeem et al. ( 2003), numerous studies have examined the associations between a uniform construct of psychopathy based on the total scores of psychopathic assessment instruments (e.g., PCL-R). This total score combines affective, interpersonal, and behavioral features of psychopathy. However, other scholars have suggested that psychopathy is a heterogeneous construct co mprised of several specific clusters and variants of psychopathic f eatures (see Cooke & Michie, 2001; Hare & Neumann, 2005; Harpur et al ., 1988, for different factor m odels of psychopathy; see

PAGE 114

103 Skeem et al., 2003, for a discussion of varian ts of psychopathy). These factors and variants have both unique and common asso ciations with other constructs (e.g., emotionality, personality traits, crime), which may reflect differences in etiology (Skeem et al., 2003). Overall, vari ants related to the behavior features of psychopathy are classified by characteristics of aggressi on, impulsivity, thrill-seeking, hostility, and low self-control. These features are simila r to the qualities of NEM and low CON and represent those qualities most likely to predispose indivi duals to delinquent (including drug use) coping in strainful situations. On the other ha nd, the variants of psychopathy that are related to affect and interpersonal ch aracteristics (e.g., lack of empathy, glibness) are marked by a quality of indifference and are least likely to lead to delinquent coping for strain. Therefore, the present study will be limited to an examination of the behavioral factor for its measur e of psychopathy, specifically the impulsivity/irresponsibility domains of psychopathy. The models described in Figures 2 and 3 are tested utilizing measures from two separate adolescent psychopathic screening devices, the An tisocial Process Screening Device (APSD) and the Youth Psychopathic feat ures Inventory (YPI). Both devices contain three factors of psychopathy (APSD : narcissism, impulsivity, and callousunemotional; YPI: grandiose-manipulative, impulsive-irresponsible, and callousunemotional). However, the models tested in this study will only include the behavioral factor of each psychopathic assessment in strument: Impulsivity for the APSD and Impulsive-Irresponsible for the YPI. Research on externalizing and internaliz ing behaviors has also demonstrated an association with aggression, delinquency, and drug use (e.g., Lynam, 2002b; Lynam et

PAGE 115

104 al., in press; Salekin, Leistico, Trobst, Schr um, & Lochman, in press). Characteristics that are indicative of externalizing be haviors, such as impulsivity, angry temperaments/dispositions, and rebelliousne ss, have also been associated with psychopathy (Brandt et al., 1997; Campbell et al., 2004; Frick, 2000; Frick et al., 2000; Hume et al., 1996; Lynam, 1997; Myers et al., 1995; Piatigorsky & Hinshaw, 2004). Characteristics that are indicat ive of internalizing behavior s, such as depression, have been associated with psychopathy, although less consistently (Brandt et al., 1997; Lynam, 1997; but see Campbell et al., 2004). Externa lizing and internalizing behaviors are also expected to influence delinquent coping in strainful situations. Figure 2 and Figure 3 are listed below. The models are identical with one exception: Figure 2 pertains to self-reported delinquency, while Figure 3 pertains to drug use problems. The following is a list of hypot heses being examined in this study by these specific models within a general strain theory framework: H1. Strain at Time 1 will be positivel y related to delinquency and drug problems at Time 2. These effects will remain significant when low social control and differential association are included in the models. H2. Strain will have positive reciprocal effects with psychopathic features of impulsivity, externalizing behavi ors, and internalizing behavior s at both Time 1 and Time 2. H3. Externalizing behavior s, internalizing behaviors, and psychopathic features of impulsivity at Time 1 will be posit ively related to strain at Time 2. H4. Strain at Time 1 will be positivel y related to externalizing behaviors and internalizing behaviors at Time 2.

PAGE 116

105 H5. Time 1 psychopathic features of im pulsivity, externalizing behaviors, and internalizing behaviors will moderate the eff ects of strain at Time 1 on delinquency and drug problems at time 2. H6. Psychopathic features of impulsivity and externalizing be haviors at Time 1 will be positively related to delinquency and dr ug problems at Time 2; while internalizing behaviors at Time 1 will be negatively rela ted to delinquency and positively related to drug problems at Time 2. Figures 2 and 3 provide an in terpretation of this struct ural model for delinquency and drug problems, respectively. Rectangles represent the following observed variables: psychopathy impulsivity, externalizing beha viors, internalizin g behaviors, and delinquency. Circles represent the various late nt variables: strain, low social control, delinquent peer associations, and drug problems. Single-headed arrows indicate the hypothesized relationships being examined, with the arrow pointing th e direction of the relationship. A double-headed arrow illustra tes the expected correlation between two variables. Thicker lines indicate which variables are hypothesized to have moderating effects leading to delinquency or drug us e problems. Non-recursive or reciprocal relationships are illustrated by the upward and downward poin ting arrows between strain and the three personality trait measures.

PAGE 117

106 Figure 2: Non-Recursive Mode l of Strain and Personality Features on Delinquency Low Social Control 1 Delinquent Peers 1 Strain1 Externalizing1 Internalizing1 Delinquency1 Delinquency2 Delinquent Peers 2 Low Social Control 2 Internalizing2 Externalizing2 Strain2 Psychopathy Impulsivity1

PAGE 118

107 Figure 3: Non-Recursive Model of Strain and Personality Features on Drug Use Problems Low Social Control 1 Delinquent Peers 1 Strain1 Externalizing1 Internalizing1 Delinquent Peers 2 Low Social Control 2 Internalizing2 Externalizing2 Strain2 Drug Use Problems 1 Drug Use Problems 2 Psychopathy Impulsivity1

PAGE 119

108 The proposed study will contribut e to the literature in several ways. First, it will serve to replicate Agnew et al.’s (2002) personality trait extension of GST; thus, potentially strengthening or weakening thei r argument. Second, Agnew and associates (2002) argue that “…much recent work in psyc hology suggests that pe rsonality traits may have a fundamental effect on the experience of and reaction to strain. In particular, the impact of such traits may be far more perv asive than that of the conditioning variables typically examined in the research.” (p. 45). Therefore, the proposed study may help to clarify the importance of personality traits as conditioning variables in the straindelinquency relationship. Furt her, the present study expands on Agnew’s personality and GST study by examining alternative and separa te measures of pers onality traits and psychopathic features, rather than a single i ndex of negative emotionality/low constraint. The present study also expands on Agnew’s pr evious study by examining the effects of personality and psychopathic features on deli nquency as well as substance use; thus providing elucidation of which personality features are more conducive to delinquent versus substance use behaviors. In additi on, this study expands on Agnew et al.’s (2002) study by examining reciprocal effects between st rain and personality. Finally, since this study examines personality characteristics, it may offer a more practical benefit by providing insight into how ma ladaptive personality charact eristics influence delinquency and drug use problems. These results may he lp to guide future policy and treatment programs that could be designed to better serve at-risk youths.

PAGE 120

109 The next chapter describes the sample used in this study. In addition, the methodology used to derive the measures used in Figure 2 and 3 are discussed. Tables presenting descriptive statistic s and other important informa tion about these measures are included.

PAGE 121

110 Chapter 5 Method A large portion of empirical tests of gene ral strain theory have utilized crosssectional data or taken a cro ss-sectional approach with long itudinal data. Overall, crosssectional studies of the relationship between st rain and delinquency have been considered methodologically appropriate because GST postu lates a contemporaneous or short-term influence of strain on deviance (Agnew, 1992). Hoffmann and Miller (1998), however, have argued that the operationa lization of strain may incl ude variables (e.g., negative relations with parents/teachers ) that cause the temporal or der of the strain-delinquency relationship to become suspect. That is, cr oss-sectional te sts of GST include variables that may be a negative outcome, rather than cause, of delinquency. For example, a youth may commit a delinquent act that his/her pare nts and teacher may fi nd out about. As a result, his/her parents and teachers may begi n to treat him/her differently, such as increasing restrictions and supervision. In this case, delinquency actually led to an increase in strain for the youth. Crosssectional tests incl uding these negative consequences can produce misleading findings that they “lead” to delinquency, when the temporal direction of the re lationship may be quite the opposite. Therefore, it is important for GST studies examining certain strain variables, such as negative parent/teacher relations, to conduct longitudinal analyses. The present study provides a

PAGE 122

111 prospective, longitudinal analysis of GST, which allows for a better examination of the temporal order of strain, delinquency, and certain moderating and mediating factors. Sample The data used in this study are secondar y data obtained as part of an innovative intervention program (called the Arbitration Intervention Worker service—see Poythress, Dembo, & DuDell, 2004) for justice referred youths in Hillsborough County, Florida. The data set contains deidentified in formation for 137 youths who voluntarily participated in the intervention and complete d both baseline and fo llow-up interviews. The Arbitration Intervention Workers Service (AIW) The AIW project was an experimental, pros pective clinical trial that evaluated the performance outcomes (e.g., program completion, r ecidivism rates, cost -effectiveness) of an intervention service invol ving arrested youths referred to a court-based juvenile diversion program, the Juvenile Arbitration di version program (referred to henceforth as Arbitration). All youths entering the Arb itration program between June 2002 and June 2003, between the ages of 11 and 18, living within a fifteen-mile radius of the Hillsborough County Juvenile Assessment Center6 (JAC) were eligible to participate in the AIW project. The Arbitration program is a juvenile diversion program that provides an alternative to adjudication for youths who have been arrested, or in some cases simply charged without being taken into custody, for a minor offense (e.g., petit theft, simple assault, disturbing the peace). Qualified yout hs are referred to the Arbitration program 6 The JAC is a centrally located, multi-agency facility designed to process all juveniles taken into custody by the various county-wide law enfor cement agencies. Juveniles are direct ed to qualifying intervention and treatment programs and detained on-site wh en necessary. (See Dembo & Brown, 1994)

PAGE 123

112 by the State Attorney’s Office. If a youth c hooses not to enter the program, he/she may be prosecuted to the full extent of the law. Youths that choose to enter Arbitration are assigned a counselor (arbitrator) who desi gnates a set of mandatory sanctions and monitors the youth’s compliance with the pr ogram. Sanctions may include restitution (e.g., community service, financial restitut ion to victim, letter of apology, etc.) and participation in psychoeducational interven tions (i.e., offense specific treatment or educational programs). Youths participate in the Arbitration program for at least five weeks. The duration of their involvement may be extended beyond the minimum five weeks from six months to a year, de pending on the curriculum of assigned psychoeducational or clinical interventions. Youths who satisfactorily complete all assigned sanctions graduate from the progr am and are spared adjudication for their offenses. All youths (and their families) volunteer ing to participate in the AIW project remained involved with the Arbitration program, receiving the usual services. Subsequent to completing a baseline intervie w, youths were randomly assigned to either the control group or the intervention group. Th e control group families were provided a telephone number that reached AIW staff to be accessed, at their discretion, should they need assistance locating local community res ources (e.g., educational programs, agencies providing financial assistance, private psychi atric and drug treatment programs, etc.). (During the AIW project, 30 families u tilized this referral service.) The intervention group received in-home, c linically supervised, case management services for a maximum duration of sixteen w eeks. The case management services were modeled after the Treatment Accountability for Safer Communities (TASC: Cook, 1992),

PAGE 124

113 which has been shown to be an effective in tervention for substance-involved youths (see Aledort, 2001; Cook, 1992, 2002; Godley et al ., 2000). The case management services were designed to (a) assist the youth in su ccessfully completing Arbitration and (b) assist families in identifying problem areas and access ing community resources best suited to assist them in dealing with any emerging problems. Previous studies of the effectiveness of the AIW intervention with respect to program compliance, drug use, recidivism, a nd other psychosocial functioning outcomes have revealed weak to non-significant treat ment effects (Dembo, Wa reham, Poythress, Cook, & Schmeidler, 2004a, 2004b, 2004c). Among the 137 youths completing both a baseline and follow-up interview, there we re no significant diffe rences between AIW intervention and non-interventi on groups for the following de mographic characteristics: age (F = 0.103, df = 1, 135, p = 0.749), sex (P earson chi-square = 0.378, p = 0.329), race (Fisher’s Exact Test = 3.950, p = 0.398), and ethnicity (Pearson chi-square = 1.514, p = 0.150). In addition, examination of differen ces (age, sex, race, ethnicity, and charges leading to placement in Arbitration) be tween youths who completed a follow-up interview (n=137) and those who did not comp lete the follow-up interview (n=28) have revealed no significant differences between th e two group on any of these characteristics (Dembo et al., under review). Hence, contro l and intervention group data were combined for this study. The sample contains only youths from a court-referred diversion program population. Although the exclus ively offending nature of the sample may lead to criticism regarding the generali zability of the results of this study, other published tests of GST have utilized offender populations (see Piquero & Sealock, 2000, 2004). According

PAGE 125

114 to Piquero and Sealock (2000, p. 454), “Insofar as GST is a general theory of criminal behavior, its applicability to offending popul ations warrants empirical study.” Other scholars have advocated the value in test ing theories utilizi ng various types of populations (Broidy & Agnew, 1997; Nagin & Paternoster, 1991). Further, some scholars (Piliavin, Thornton, Gartner, & Ma tsueda, 1986, p. 104) have suggested that research focused on criminal offenders and offenses are benefici al for public policy decisions. Therefore, similar to Piquero and Sealock, this paper attempts to investigate the “generality” of general strain theory in characterizing deli nquency in an offending population. Unlike the sample of juvenile s placed on probation that comprised the Piquero and Sealock (2001, 2004) samples, the youths examined in this study are mostly first-time misdemeanor offenders. In this regard, the present study may benefit not only policy-makers, in general, but also those inte rested in juvenile intervention, treatment, and prevention. Sociodemographic Information at Time 1 Table 1 contains sociodemographic in formation about the youths and their families collected during the baseline interv iew. These measures are similar to the sociodemographic information reported by Ag new et al. (2002). The sample is comprised of slightly more male adolescents (51.8%) than female adolescents (48.2%). Most of the youths identified themselves as White (63.5%); approximately 36 percent considered their racial identi ty to be something other th an White (most were AfricanAmerican). Regardless of race, approxima tely 25 percent of the youths considered themselves to be of Hispanic ethnicity. Th e youths ranged in age from 11 to 18 years old

PAGE 126

115 (the age range reflects the eligibility crit eria discussed above). The average age of participants was 14 (standard deviation = 1.697). A majority of youths (77.4%) lived in fam ily situations with only one biological parent (e.g., single parent, rema rried parent, single parent liv ing with a significant other but not married). Approximately 20 percent still lived with both of their biological parents; 3 percent lived with neither biological parent. Yout hs were asked to indicate the total number of years of education each pa rent/guardian had received. Regarding the educational background of primary caretakers, 39 percent had rece ived 12 years of school, and almost 35 percent had a ttended school beyond high school. Based on a hierarchy of occupational status developed by Hollingshead and Redlich (1958), a proxy measure of socio economic status (SES) was created using information pertaining to the occupation of the head of th e household. The distribution of the occupational status variable suggests th at the youths in this study lived in families with low to moderate SES. Only 6 percent of the chief wage earners in these families held higher executive, administrative, or ma nagement positions. Twenty-seven percent of the chief wage earners held skilled or semi-skilled positions, while 13 percent held unskilled positions or were unemployed. Seven percent of the youths could not describe or did not know what type of job the head of their household held.

PAGE 127

116 Table 1: Sociodemographic Information at Time of Baseline Interview (N = 137) Gender n % Race n % Female 66 48.2 Non-White 50 36.5 Male 71 51.8 White 87 63.5 137 100.0 137 100.0 Living with n % Ethnicity n % Both Biological Parents 27 19.7 Hispanic 35 25.5 Neither Parent 4 2.9 Non-Hispanic 102 74.5 Single Biological Parent 106 77.4 137 100.0 137 100.0 Age n % Years of Education for Primary Parent n % 11 3 2.2 3 1 0.7 12 22 16.1 8 1 0.7 13 22 16.1 9 4 2.9 14 28 20.4 10 5 3.6 15 22 16.1 11 7 5.1 16 24 17.5 12 53 38.7 17 15 10.9 13 8 5.8 18 1 0.7 14 15 10.9 137 100.0 15 4 2.9 16 17 12.4 Mean = 14.32 17 1 0.7 Standard Deviation = 1.697 18 2 1.5 20 1 0.7 Unknown/Missing 18 13.1 137 100.0 (Continued on the next page)

PAGE 128

117 Table 1: (Continued) Family Occupation Level n % Medium business managers and lesser professionals 8 5.8 Administrative personnel, managers, minor professionals and small business owners 15 10.9 Clerical and sales, technicians and sm all businesses 50 36.5 Skilled manual labor 18 13.1 Semi-skilled 19 13.9 Unskilled & unemployed 18 13.1 Unknown or uncodable information 9 6.7 137 100.0 Measures The measures described below were collect ed as part of a baseline and follow-up protocol that was administered to youths particip ating in the AIW clinic trial described above. The majority of this protocol included manual administration of the CASI (including the CASI addendum questions, usua lly not included in the computerized version) (Meyers et al., 1999). Youths were also administered the Antisocial Process Screening Device (APSD) (Frick & Hare, 2001), the Youth Psychopathic features Inventory (YPI) (Andershed, Ke rr, Stattin, & Levander, 20 02), and the National Youth Survey (NYS) (Elliott, Ageton, Huizinga, Know les, & Cantor, 1983) as part of the AIW project protocol.

PAGE 129

118 The Comprehensive Adolescent Severity Inventory (CASI) The CASI (Meyers et al., 1999) is a se mi-structured, clinical assessment and outcomes instrument that collects information on youths’ psychosocial problems and strength-resiliency factors acr oss a number of life areas. It is comprised of ten independent modules, each reflecting separate life areas: drug/alc ohol use; education; family/household member relationships; health information; legal issues; mental health; peer relationships; sexual behavior; stressful life events; and use of free time. (In this study the legal issues, sexual behavior, and stressful life events modules were not administered as part of the AIW project protocol.) According to Meyers and her associates (Meyers et al., 1999, p. 239), the primary life areas covered by the CASI provide a broad enough assessment of adolescent functioning that they can be used to inform clinical adolescent treatment. The CASI has demonstrated excellent psychometric properties (Meyers et al., in press; Me yers, Webb, Hagan, & Frantz, submitted). Questions contained in the CASI address whether or not certain behaviors have ever occurred, occurred within the past month occurred within the other 11 months of the past year, and the age of onset For the most part, the answers are dichotomous (1 = yes, 0 = no). The questions are phrased so that the youth only responds affirmatively if the question refers to a condition or event that has occurred for a “signi ficant period” during the appropriate time frame (e.g., past month, ot her 11 months, past year). “Significant period” is a rather ambiguous term, but in terviewers are instructed by the CASI developers and training instruct ors to inform, and regularly remind, interviewees that the term refers to anything that has occurred long enough or often enough that is has become a problem with regard to the life area being addressed. In this sens e, the CASI responses

PAGE 130

119 represent subjective evaluations of objectively defi ned life area events and conditions (see Agnew, 2001). There are, however, some exceptions where isolated incidents, rather than those occurring over a significant duration, are reco rded (rape/sexual assault, physical abuse, sexual abuse, animal cruelty, ar son/fire setting, suic ide attempts, and selfmutilation). In addition, the age of onset porti on of each question refers to the first time the youth experienced or performed a specific problematic life area event or condition. The developers of the CASI provide a sc oring manual that can be utilized to construct theoretically appropr iate and psychometrically so und subscales of risk and protective behaviors or symptoms. The s ubscales reflect raw sc ores indicating the presence or absence of certai n risk and protective behaviors. The subscales are created by taking the average of the su m of specified dichotomous variables (0 = no, 1 = yes) within each life area module. Scores for th e subscales range from 0 to 1, with a higher score indicating greater risk or protective factors, dependi ng on the subscale. The predefined subscales for externalizing behaviors, internalizing behavi ors, and alcohol/drug related problems (serious consequences, narrowi ng of behavior repertoire, loss of control, and physical dependence scales) were used in this study for the year prior to the baseline interview (Time 1) and the year prior to the follow-up interview (Time 2). The initial intention was to utilize the CASI subscales for life areas of family, education, and peer associations to replic ate the Agnew et al. (2002) study. Further examination of the items included in the CASI subscales, however, revealed considerable conceptual overlap regarding GST and social c ontrol measures within scales. Therefore, individual items were used to create more appr opriate scales for this particular sample of youths.

PAGE 131

120 Deriving Appropriate Measures from the CASI Similar to Agnew et al. (2002), items belie ved to be measures of strain, social control, or differential associ ation were grouped into three categories: family, peer, and school. Then an exploratory factor anal ysis (EFA) was conducted for theoretically relevant measures within each category (family, peer, and school, separately) for Time 1 (baseline). This approach was utilized to maintain conceptual distinction between the strain items and social control items within each life area (see Agnew et al., 2002, p. 50). Since the items from the CASI were cat egorical, each EFA was performed using Mplus version 3.12 (Muthn & Muthn, 2004). Mplus is a multivariate statistical modeling program that estimates a variety of simple and sophisticated models (e.g., path analysis, growth models, multilevel models) for continuous and categorical, observed and latent variables. In these analyses, a chi-square test is used to test the fit of the models to the data. Lack of significance indi cates an acceptable model fit. Mplus also provides a number of descrip tive fit measures to assess the closeness of fit of the model to the data. Three fit indices were used to evaluate the model fit: (1) the comparative fit index (CFI) (Bentler, 1990) (2) the Tucker-Lewis coefficient (TLI) (Tucker & Lewis, 1973), and (3) root mean square error of approximation (RMSEA) (Byrne, 2001). The typical range for both TLI and CFI is between 0 and 1 (although TLI can exceed 1.0), with values greater than .95 indicating a good fit (Browne & Cudeck, 1993; Hu & Bentler, 1999). For RMSEA, values at .05 or less indicate a close model fit, and values between .05 and .08 indicating a mediocre model fit (Browne & Cudeck, 1993). In addition, Mplus was utilized becau se it contains a missing data imputation procedure for both categorical and conti nuous variables. The missing imputation

PAGE 132

121 component was especially important in this study for preserving the sample size; thus, maintaining power in the analyses. Unlike Agnew’s test of GST and personali ty, orthogonally rotate d factor scores, rather than oblique, were used as a basis for creating the strain, social control, and differential association measures. Orthogonal rotation is “aimed at maximizing variance of the factors” (Pedhazur & Schmelki n, 1991, p. 613) and minimizing the correlation between factors. Although it is very likely that the famil y, school, and peer factors are moderately correlated, orthogonal rotation was used to minimize conceptual overlap for the factor. (The loadings for EFA promax co rrelated rotations were very similar to the loadings of the Varimax rotations. Theref ore, the decision to choose orthogonal rotation over correlated factor rotation for interpreta tion of the data did not impact measurement decisions. The promax oblique correlati on values are reported in Appendix A.) Appendix A reports the Varimax rotated EF A results for Time 1 family, peer, and school categorical items. (EFA results suppor ted a priori speculation for membership of the family, peer, and school items into factor s of strain, social control, and differential association.) For Time 1, 98.5 percent or more of the data were present for missing imputation within family, peer, and school it ems. Within the family category, twenty items were examined. For the family EFA, seven factors with an eigenvalue greater than one were identified in the data. However, the data loaded well onto three factors (Chisquare = 38.86, df = 29, p = 0.10; RMSEA = 0.050) and examination of output for four or more factors did not indicate a substant ial improvement in the fit of the data. Therefore, three factors were examined in s ubsequent confirmatory factor analyses of Time 1 and Time 2 measures.

PAGE 133

122 Within the peer category, eleven items we re examined. For the peer EFA, three factors with an eigenvalu e greater than one were identified in the data. However, the data loaded well onto two factors (Chi-squar e = 22.08, df = 17, p = 0.18; RMSEA = 0.047). Therefore, two factors were examined in s ubsequent confirmatory factor analyses of Time 1 and Time 2 measures. Within the school category, seven items were examined. For the school items EFA, two factors with an eigenvalue greater than one were identified in the data. The data loaded very well onto these two factors (Chi-squa re = 4.02, df = 6, p = 0.67; RMSEA = 0.000). If the factor results for Time 1 were rep licated in confirmato ry factor analyses (CFA) results for Time 2 measures, greater c onfidence in the validity of the factors would be established for this particular sample. Therefore, CFAs were conducted on family, peer, and school items for Time 1 and Time 2, respectively, to examin e the validity of the measures. (Mplus does not save factor scores for data using EFA; fa ctor scores can only be saved using CFA techniques.) Tables 2, 3, and 4 describe the CFA results for the individual CASI items. For the family items, CFAs specifying three factors for Time 1 and Time 2 measures were completed and found to fit the data rather well (Time 1: chi-square = 43.73, df = 34, p = 0.12; CFI = 0.952; TLI = 0.956; RMSEA = 0.046; Time 2: chi-square = 25.06, df = 17, p = 0.09; CFI = 0.942; TLI = 0.925; RMSEA = 0.059). Two factors appeared to describe strain measures of family disruption and family abuse/neglect The third factor described low so cial control measures of parental attachment and commitment The item loadings ranged from .35 to .91. Each of the variables loaded

PAGE 134

123 significantly on these factors; however, at Time 2 the “other member ignored or given the silent treatment” item was only marginally significant (critical-ratio = 1.944). Summary factor scores developed by Mplus were save d for use in the models, where higher scores indicate family related problems. For the peer items, CFAs specifying two f actors for Time 1 and Time 2 measures were completed and found to fit the data we ll (Time 1: chi-square = 19.92, df = 17, p = 0.28; CFI = 0.987; TLI = 0.988; RMSEA = 0.035; Time 2: chi-square = 20.33, df = 19, p = 0.37; CFI = 0.996; TLI = 0.996; RMSEA = 0.023). One factor reflected strain measures of negative or poor peer relationships. The ot her factors described delinquent peer associations The item loadings ranged from .49 to .94. Each of the variables loaded significantly on these factors. Summar y factor scores deve loped by Mplus were saved for use in the models, where higher sc ores indicate peer relationship problems and delinquent peer associations. For the school items, CFAs specifyin g two factors for Time 1 and Time 2 measures were completed and found to fit the data rather well (Time 1: chi-square = 6.77, df = 9, p = 0.66; CFI = 1.000; TLI = 1.017; RM SEA = 0.000; Time 2: chi-square = 7.41, df = 8, p = 0.49; CFI = 1.000; TLI = 1.007; RM SEA = 0.000). The factors described the social control measures of school attachment and school commitment Unlike Agnew et al. (2002), none of the factors described strain measures of negative school relationships. The factor loadings ranged from .39 to .98. Each of the variables loaded significantly on these factors; however, at Time 2 “felt sa fe at school” (reverse coded) was only marginally significant (critica l-ratio = 1.729). Mplus summary factor scores were saved for use in the models, where higher scores indicate school problems.

PAGE 135

124 Table 2: CFA Standardized Loadings for Family Items for Time 1 and Time 2 Time 1 Time 2 Latent Variable Family Items Standardized Loadings Standardized Loadings Family Repeatedly insulted/criticized .85 .70 Disruption Other member insulted criticized .54 .68 Ignored or given “silent treatment” .90 .58 Home felt like safe place [reversed] .59 .89 Family works out problems non-violently [reversed] .66 .78 Ran away from home .62 .47 Felt loved by someone in home [reversed] .67 .91 Family contacted about domestic disputes .48 .70 Eigenvalue = 3.66 4.25 Variance = 45.8 53.1 Parental Couldn’t get along/fighting with family member .51 .56 Attachment Parents disagree on limits/punishment .75 .60 Hard to talk to/confide in parents .76 .72 Parents don’t listen to you .83 .90 Parents unavailable to you .76 .84 Parents covered/made excuses for you .44 .40 Rules not consistently enforced .55 .56 Given praise for good behavior [reversed] .66 .68 Parents really know what/where you go/do [reversed] .35 .53 Eigenvalue = 3.73 3.93 Variance = 41.5 43.6 (Continued on the next page)

PAGE 136

125 Table 2: (Continued) Time 1 Time 2 Latent Variable Family Items Standardized Loadings Standardized Loadings Family Other member threw object, punched walls .73 .46 Abuse Hit hard (physically abused) .82 .71 Other member ignored or “silent treatment” .86 .43 Eigenvalue = 1.94 0.90 Variance = 64.8 30.1 Time 1: 2 = 43.73, df = 34, p = 0.12; CFI = 0.952; TLI = 0.956; RMSEA = 0.046. Time 2: 2 = 25.06, df = 17, p = 0.09; CFI = 0.942; TLI = 0.925; RMSEA = 0.059.

PAGE 137

126 Table 3: CFA Standardized Loadings fo r Peer Items for Time 1 and Time 2 Time 1 Time 2 Latent Variable Peer Items Standardized Loadings Standardized Loadings Peer Difficulty making/keeping friends .85 .82 Strain Had no friends .55 .64 Preferred to be alone .67 .76 Felt friends were not loyal .76 .94 Hard to talk to friends .87 .84 Dissatisfied with quality of friendships .81 .86 Consistently teased/bullied .49 .63 Eigenvalue = 3.70 4.40 Variance = 52.9 62.8 Delinquent Hung out with people who use drugs/drink .92 .88 Peers Hung out with people who commit illegal acts .74 .84 Hung out with gang members .80 .86 Hung out with people who skipped/dropped school .70 .84 Eigenvalue = 2.53 2.91 Variance = 63.3 72.7 Time 1: 2 = 19.92, df = 17, p = 0.28; CFI = 0.987; TLI = 0.988; RMSEA = 0.035. Time 2: 2 = 20.33, df = 19, p = 0.37; CFI = 0.996; TLI = 0.996; RMSEA = 0.023.

PAGE 138

127 Table 4: CFA Standardized Loadings for School Items for Time 1 and Time 2 Time 1 Time 2 Latent Variable School Items Standardized Loadings Standardized Loadings School Had failing grades/difficulty learning .71 .65 Commitment Skipped class/arrived late consistently .59 .81 Were suspended, expelled, had detention .42 .67 Had little or no interest in school .98 .83 Eigenvalue = 1.98 2.20 Variance = 49.6 55.0 School Went to school prepared [reversed] .86 .98 Attachment Felt you belonged in school [reversed] .78 .84 Felt safe at school [reversed] .60 .39 Eigenvalue = 1.70 1.82 Variance = 56.8 60.8 Time 1: 2 = 6.77, df = 9, p = 0.66; CFI = 1.000; TLI = 1.017; RMSEA = 0.000. Time 2: 2 = 7.41, df = 8, p = 0.49; CFI = 1.000; TLI = 1.007; RMSEA = 0.000.

PAGE 139

128 Strain Measures Three measures of strain were estimate d as comprising one latent variable of strain for Time 1 and Time 2: family disruption, family abuse, and poor peer relationships. These measures are similar in nature to those incl uded in Agnew et al. (2002), except without the school strain m easure. One potential weakness of the measures in this study, however, is that multi-informants were not available for the sample being studied. The items reflect yout h self-report data only. The strain items examined in this study, however, may repr esent more subjective appraisals of each measure because youths were directed to provi de affirmative responses only if the event or condition occurred for a problematic period of time (see Agnew, 2001 for a discussion). Family disruption Youths were asked to indicate whether several statements pertaining to family or house hold (i.e., any person living in the residence with them) relationships had occurred for significant pe riods during the past year. For all CASI items, responses for each statement were dichotomous (1 = yes, 0 = no). Past year measures were created for Time 1 by combining “past month” and past “other 11 months” responses. Past year measures for Time 2 reflected responses from the “since last contact” items. (For detailed discu ssion of CASI item res ponse categories see the above mentioned section describing the CASI.) Therefore, possible responses for Time 1 were 0 = no significant problems, 1 = significant problem in past month or past other 11 months, or 2 = significant problems in both past month and other 11 months, and possible responses for Time 2 were 0 = no significan t problems or 1 = significant problems since last contact. High scorers on th e family disruption fact or (see Table 2) stated that in their

PAGE 140

129 family life they were “repeatedly insulted/ criticized,” “ignored or given the silent treatment,” other family members were “repeatedly insulted/criticized,” the family was “contacted by police… about domestic disputes,” home did not feel “like a safe place,” the family could not “work out problems with you in a non-violent manner,” they did not feel “loved by someone” in the family, and they “ran away from home.” In short, high scorers on this measure experien ced disruptive, potentially viol ent, negative relationships with their family members (alpha reli ability: Time 1 =.72, Time 2 = .56). Family abuse/neglect Youths were asked to indi cate whether three statements pertaining to family or household abuse or neglect had occurred for significant periods during the past year. Similar to the family disruption items, responses for these items were also categorical, ranging fr om 0 to 2 for Time 1 and 0 to 1 for Time 2 past year occurrences. High scorers on the family abuse f actor stated that in their family life they were “hit so hard they had… bruises, br oken bones…,” other family members “threw objects, punched walls…” when angry, and other family members were “ignored” for extended periods of time. In short, high scor ers on this measure experienced abusive or neglectful family relationships (alpha reliab ility: Time 1 = .64, Ti me 2 = -.09). Although the CFA for Time 2 suggested a good fit of the family factors to the model, the internal consistency of the family abuse factor for Time 2 is very weak. Indeed, it seems the Time 2 family abuse scale items are not correl ated. As seen in Table 2, the factor does not exceed an eigenvalue of 1, which also sugge sts less coherence within this factor. The lack of coherence over time proved problema tic for the SEM analyses discussed in Chapter 6.

PAGE 141

130 Peer strain. This scale contains 7 items fr om the CASI peer relationships module. Youths were asked to indicate whether statements re ferring to poor peer relationships had occurred for significant peri ods during the past year. Similar to the family disruption items, responses for these items were also categorical, ranging from 0 to 2 for Time 1 and 0 to 1 for Time 2. High scorers on the peer strain factor stated that among their friends and peers they “preferred to be alone,” “had no friends,” felt their friends were “not loyal,” “ had “difficulty making/keeping friends,” found it “hard to talk to …” their friends, “were “dissatisfied with the quality of ...friendships,” and were “consistently teased or bullied by…peers.” High scorers on this measure experienced negative relationships with peers and friends (alpha reliability: Time 1 = .76, Time 2 = .78). Social Control Measures Three measures of social control were es timated as comprising one latent variable for low social control for Time 1 and follo w-up Time 2, respectively, (see Tables 2 and 4). These measures are also similar in nature to those included in Agnew et al.’s (2002) test of GST and personality traits. Thes e measures refer to proximal relationships between the youth and his/her family and di stal relationships between the youth and school. Low parental attachment/commitment This scale contains nine items from the CASI family/household relations hips module. Youths were asked to indicate whether statements referring to familial relationships had occurred for significant periods during the past year. Similar to th e strain items, responses for these items were categorical, ranging from 0 to 2 for Time 1 and 0 to 1 for Time 2. High scorers on the parental

PAGE 142

131 attachment/commitment factor stated that they “coul dn’t get along with another household member,” “found it hard to…talk wi th” their parents, “rules were not consistently enforced,” were not “given credit or praise fo r doing the right thing,” their parents “disagreed on…what lim its…consequences” to set, “did not listen to what [they] had to say,” “were unavailabl e to” them, “covered for” th em, and did not “really know where…who [they] hung out w ith.” High scorers on this measure experienced low parental attachment/commitment (alpha re liability: Time 1 = .73, Time 2 = .66). Low school attachment Responses to three items in the CASI education module loaded highly on this factor. Youths were asked to indicate whet her statements referring to school experiences had occu rred for significant periods dur ing the past year. Similar to the strain items, responses for these items were categorical, ranging from 0 to 2 for Time 1 and 0 to 1 for Time 2. High scorers on the school attachment factor stated that they did not go to “school prepared,” did not feel they “belonged in school,” and did not feel “safe at school.” High scorers on this measure experienced low school attachment (alpha reliability: Time 1 = .60, Time 2 = .49). Low school commitment Responses to four items in the CASI education module loaded highly on this factor. Youths were asked to indicate whet her statements referring to school experiences had occu rred for significant periods dur ing the past year. Similar to the strain items, responses for these items were categorical, ranging from 0 to 2 for Time 1 and 0 to 1 for Time 2. High scorers on the school commitment factor stated that they “had failing grades or had difficulty learning,” “cut class/school…on a consistent basis,” were “suspended, expelled, had nume rous detentions,” and “had little or no

PAGE 143

132 interest in school.” High scorers on this measure experi enced low school commitment (alpha reliability: Time 1 = .63, Time 2 = .66). At the time of the baseline interview, nine youths were not actively in school. Seven youths had been out of school less th an 12 months, one of whom had graduated from high school. One youth was being home schooled, and had been doing so for the past 3 years. One youth had been out of sc hool for almost 2 years. For the baseline CASI, youths not attending school were aske d the exact same questions as those attending school. When responding to the ba seline education questions, however, youths not currently enrolled in school were asked to refer to the 12 months prior to leaving school. For youths not attending school at Time 1, comparable items referring to the 12 months prior to thei r last day in school were used. For the follow-up interview, the CASI does not contain items for youths not attending school that are comparable to items provided for youths attending school. There were eighteen youths not attending school during Time 2. Missing imputations were utilized to create the la tent factor for social contro l at Time 2 for missing cases on the school attachment and school commitment measures. Social Learning/Differential Association Measures One measure of differential association was examined for both Time 1 and Time 2. Agnew et al.’s (2002) test of GST and personality tr aits included a single item measuring parental perceptions of their children’ s delinquent peers. The present study examines a factor of differential associat ion that contains four items concerning delinquent peers. Youths were asked to i ndicate whether or not they hung around people who “used drugs or got drunk regularly,” “comm itted illegal acts,” “were members of a

PAGE 144

133 gang,” and “dropped out of sc hool…didn’t attend regularly.” As with the strain and social control items, responses for these item s were categorical, ranging from 0 to 2 for Time 1 and 0 to 1 for Time 2. High scorers on this measure were highly associated with delinquent/deviant peers (alpha reli ability: Time 1 = .75, Time 2 = .76). Personality and Psychopathic Features The present study examined the infl uence of maladaptive personality characteristics using psychopathy, externalizing behaviors, and intern alizing behaviors as proxy measures. As previously mentioned, research has suggeste d the construct of psychopathy may be heterogeneous (Karpm an, 1941, 1948; Porter, 1996; Mealey, 1995a, 1995b). To reiterate, psychopathy may be be st characterized as having two variants: primary and secondary psychopathy. Primary psychopathy is believed to be caused by genetic factors and motivated by purposeful effo rts to satisfy desires. This variant of psychopathy taps the affective dimension of the construct. In contrast, secondary psychopathy is believed to be caused by negative psychosocial and environmental conditions (e.g., abuse, parent al rejection/neglect) (Forth & Burke, 1998; Margolin & Gordis, 2000; Marshall & Cooke, 1995; Porter, 1996; Weiler & Widom, 1996) and motivated by emotional and impulsive responses to negative environmental circumstances. This variant of psychopat hy taps the behavioral dimension of the construct. As such, the beha vioral domain of psychopathy seems more in line with the theoretical premise of GST. Therefore, th e models examined in the present study are limited to the inclusion of the impulsivity/i rresponsibility or behavioral domains of psychopathy.

PAGE 145

134 This study used two measures of ps ychopathy: (1) the Antisocial Process Screening Device (APSD: Frick & Hare, 2001) and (2) the Youth Psychopathic traits Inventory (YPI: Andershed et al., 2002). In addition, proxy measures for personality disposition were examined using the CASI subscales for externalizing behavior and internalizing behavior. Criminological tests of persona lity, especially antisocial personality disorder and psychopathy, must be cautious when c onstructing measures. Since many of the assessments for these types of maladaptive be haviors include measures of criminality as part of their assessment, it is essential to eliminate any poten tial criterion contamination. One item of the APSD refers to criminal activity, which creates criterion contamination when attempting to study the causation of crime. This item, however, does not contribute to the impulsivity or behavioral domain be ing studied here. The YPI does not contain items measuring criminal activity. APSD psychopathic features The self-report version of the APSD contains 20 items that measure several of the same psychopathic features as the Psychopathic Checklist-Revised (Hare, 1991) (discussed in chapter 3). The APSD was initially designed for use with youths between the ages of 6 and 13 years old with ratings provided by familiar adults (e.g., parents, teachers, etc.), rather than as a self-report tool. The APSD items yield a Total score of psyc hopathic features and load well onto a threefactor structure for describing psychopathic features: (1) Narcissism, an interpersonal factor; (2) Callous-Unemotional, an affective factor; and (3) Impulsive, a behavioral factor (Frick et al., 2000). Research supports the construct validity of the administered

PAGE 146

135 version of the APSD (see Blair, 1999; Bl air, Monson, & Frederickson, 2001; Loney, Frick, Clements, Ellis, & Kerlin, 2003; O’Brien & Frick, 1996). Although the APSD was not sp ecifically designed for us e with justice-involved youths, a previous analysis of youths fr om the AIW project (Poythress, Dembo, Wareham, & Greenbaum, in press) revealed the APSD possessed adequate psychometric properties. In addition, limited research ha s revealed promising results regarding the construct validity of th e self-report APSD, thou gh the internal consistency of the factors have been modest (Falkenbach et al., 2003; Lee et al., 2003; Murrie & Cornell, 2002; Pardini et al., 2003). The self-report version of the APSD is designed for use with older adolescents (i.e., between 12 and 18 years of age). The se lf-report APSD items yield a Total score of psychopathic features and load onto the three-factor structur e for describing psychopathic features (Vitacco et al., 2003) Narcissism is comprised of 7 items, such as “Your emotions are shallow and fake.” Callous-Unemotional contains 6 items, such as “You are concerned about the feelings of others (r everse scored).” Impul sive includes 5 items, such as “You act without thinking of the cons equences.” Each item is rated on a 3-point scale, with responses indicating not at all true (0), sometimes true (1), or definitely true (2). The APSD was only administered once dur ing the AIW clinical trial, at Time 1. A summary score was created for the A PSD impulsivity domain. This additive index contained the five APSD impulsivity domain items. Possible scores ranged from 0 to 10. High scorers on the APSD impulsivity index indicated higher impulsive and risktaking characteristics (a lpha reliability = .54).

PAGE 147

136 YPI psychopathic features The YPI is a 50-item measure containing items that represent several of the psyc hopathic dimensions of the Ps ychopathic Checklist-Revised (Hare, 1991) as well. The YPI is a self-report tool that was designed to be administered to youths over the age of 12. According to its developers (Ande rshed et al., 2002), the YPI offers an advantage over the APSD becau se it was designed specifically for selfreport and may suffer from less response bi as based on the phrasing of the YPI items versus the APSD items. The YPI also contains multiple items to represent the Psychopathic Checklist-Revised psychopathic features (see Falkenbach et al., 2003). The YPI items yield a Total score of ps ychopathic features and load well onto a hierarchical three-factor m odel of personality traits (c f. Cooke & Michie, 2001): (1) Grandiose-Manipulative, an interpersonal fa ctor; (2) Callous-Unemotional, an affective factor; and (3) Impulsive-Irres ponsible, a behavioral factor. Each factor or domain contains multiple sub-scales (a total of 10) of psychopathic features. Within the Grandiose-Manipulative domain, sub-scales are created for Dishonest Charm (e.g., “It’s easy for me to charm and se duce other to get what I want from them”), Grandiosity (e.g., “I have tale nts that go far beyond other pe ople’s”), Manipulation (e.g., “I am good at getting people to believe in me when I make something up”), and Lying (e.g., “Sometimes I lie for no reason, other than because it is fun”). Within the CallousUnemotional domain, sub-scales character ize Remorselessness (e.g., “When someone finds out about something that I’ve done wrong, I feel more angry than guilt”), Unemotionality (e.g., “To be nervous and worried is a sign of weakness”), and Callousness (e.g., “I think that crying is a si gn of weakness, even if no one sees you”). Within the Impulsive-Irresponsible domain, s ub-scales describe for Thrill-Seeking (e.g.,

PAGE 148

137 “I like to be where exciting thing happen”), Impulsivity (e.g., “It often happens that I talk first and think later”), and Irre sponsibility (e.g., “If I won a lo t of money in the lottery I would quit school or work and just do things that are fun”). Respondents are asked to rate the degree to which each item applies to them using a 4-point Likert-type response: does not apply at all (0), does not apply well (1), applies fairly well (2), or applies very well (3). The YPI was only administered once during the AIW clinical trial, at Time 1. The YPI was also not specifically designed for use with justice-involved youths. Skeem and Cauffman (2003) have publishe d one of the only studies examining the application of the YPI in a delinquent sample. Their resu lts suggest the YPI possesses adequate internal consistency and concu rrent validity when compared to the youth version of the Psychopathic Checklist-Revised. Summary scores were created for the YPI impulsivity-irresponsibility domain, as well as the three dimensions within this dom ain—impulsivity, irresponsibility, and thrillseeking. The additive index for the YPI im pulsivity-irresponsibility scale contained 15 items and 5 items for each dimension. Possible scores for the overall domain ranged from 0 to 45. Scores for the three dime nsions comprising the overall impulsivityirresponsibility domain range from 0 to 15. High scorers on the YPI impulsivity index reported higher impulsive, irresponsible, and ri sk-taking characteristic s (alpha reliability: Impulsivity-irresponsibility = .84, Impulsivity = .68, Irresponsi bility = .68, Thrill-seeking = .69). CASI mental health measures In addition to the psychopathy measures, two indexes from the CASI mental health module were included in this study: externalizing problems and internalizing problems. The ex ternalizing problems indexes past year

PAGE 149

138 responses to five items that describe various behavioral c onstructs, such as hyperactivity, impulsivity, thrill-seekin g, rebelliousness, and hostility. Youths were asked to indicate whether or not (0 = no, 1 = yes) they had expe rienced significant ment al health life area problems over the past year. These items refe rred to externalizing behavioral problems such as “restless, fidgety…extremely distract ible,” “impulsive, did dangerous things for the thrill of it,” “intention ally violate rules,” “consiste ntly lost your temper,” and “extremely hostile…excessive outbursts.” Responses for these items were summarized as an additive index, ranging from 0 to 10 fo r Time 1 (for Time 1 combined ten items: past month and past other 11 months) and 0 to 5 for Time 2. High scorers on this measure indicated expressing greater amount s of externalizing behaviors (alpha reliability: Time 1 = .82, Time 2 = .68). The Internalizing problems indexes were comprised of past year responses to seven items that describe vari ous behavioral constructs, such as depression, anxiety, and low self-esteem. Youths were asked to indica te whether or not (0 = no, 1 = yes) they had experienced significant mental health life area problems over the past year. These items referred to internalizing behavioral problems such as “thoughts of failure, lacked self confidence,” “extremely intimidated, shy,” “ex tremely anxious, felt panicky,” “constantly preoccupied with food,” “thoughts you could not get rid of…same things over and over again,” “sad, hopeless…cried a lo t,” “extremely tired or had l ittle energy.” Responses for these items were summarized as an additive index, ranging from 0 to 14 for Time 1 (for Time 1 combined fourteen items: past mont h and past other 11 months) and 0 to 7 for Time 2. High scorers on this measure expr essed greater problems with internalizing behaviors (alpha reliability: Time 1 =.86, Time 2 = .79).

PAGE 150

139 Delinquency Youths in this study provided self-report delinquency information as part of the baseline and follow-up protocol. The self-r eport delinquency measur es were based on the work of Elliott and associates (1983). Respondents were asked to indicate how many times within the past 12 mont hs they had engaged in 23 specific delinquent behaviors (e.g., stole a motor vehicle, stole someth ing worth between $5 and $50, broke into a building or vehicle, aggravated assault, hi t a student, etc.). Youths who reported committing an offense 10 or more times within the past year were also asked to provide an average estimate of how often they enga ged in such behavior based on specific frequency categories (i.e., two or more ti mes a day, once a day, two or three times a week, once a week, once every tw o or three weeks, once a month). This served as a check-and-balance system to cont rol overestimation. Youths we re also asked to indicate the age of onset for any behavior they admitted to committing. Based on this information, a summary meas ure of 18 of the NYS items for selfreported delinquency was created. Youths we re asked how many times within the past 12 months they had committed the following: stolen a motor vehicle; gone joyriding; broke into a building or vehicl e; stolen something worth more than $50; stolen something worth between $5 and $50; stolen something wo rth $5 or less; used force to get money from a student; used force to get money from a teacher; used force to get money from other people; held stolen goods ; carried a hidden weapon; atta cked someone with the idea of hurting them; been paid for having sexual re lations; had sexual relations with someone against their will; been involved in a gang fi ght; hit a teacher; hit a parent; and hit a student. Responses for each of these questions were summed to create an additive scale

PAGE 151

140 of delinquency. High scorers on this measure admitted engaging in more delinquent behavior (alpha reliability: Time 1 = .29, Time 2 = .43; alpha reliability (log transformed): Time 1 = .67, Time 2 = .79). The following five items were omitted from the total delinquency index: sold marijuana/hashish; sold cocaine/crack; sold other hard drugs; been loud/rowdy in a public place; and begged for money from strangers. The drug items were omitted because the CASI drug problems measure (described belo w) captured legal problems with drugs. The other two items were excluded because th ey referred to less severe delinquent/public nuisance acts and were not consistent with the severity of the other 18 items. Table 5 indicates that the 137 youths repor ted relatively high rates of delinquency. As illustrated, there is a high prevalence rate for total delinquency (Time 1: 80%, Time 2: 41%), with 2 percent and 6 percent for Time 1 and Time 2, respectively, of the adolescents reporting engagement in 100 times or more of these offenses. Since the delinquency scale reflected high skewness (w ithout log transformation: Time 1 = 6.89, Time 2 = 6.50; log transformed: Time 1 = 0.09, Time 2 = 1.33) and kurtosis (without log transformation: Time 1 = 54.04, Time 2 = 48.28; log transformed: Time 1 = -0.19, Time 2 = 0.98), it was logarithmically transformed to the base 10, with –1 being assigned to students reporting no delinquent offenses (after taking the l og). This scoring provided a more meaningful interpretation of differences in terms of delinquent involvement than the raw scores. That is, equal intervals on the transformed scale would represent equal differences in involvement. Specifically, the differences be tween no offense and 1, 1 and 10, 10 and 100, and 100 and 1000 offenses would be interpreted as comparable.

PAGE 152

141 Table 5: Self-Reporte d Delinquency (N = 137) Frequency during 12 Months Prior to Baseline Interview 0 1-4 5-29 30-54 55-99 100-199 200+ Total Total Delinquency 20% 50% 23% 4% 0% <1% 2% 100% Frequency during Period between Baseline and Follow-Up Interview 0 1-4 5-29 30-54 55-99 100-199 200+ Total Total Delinquency 59% 24% 13% 1% 0% 2% 4% 100% Drug and Alcohol Use Problems Research has shown that substance use and abuse among youths can be described as a series of progressively more intens e or severe usage (Kandel, 1975, 2002; Kandel, Kessler, & Margulies, 1978; Kandel, Yama guchi, & Chen, 1992). Most youths begin their substance use with tobacco or alcohol us e, which progresses to marijuana, and then the use of other drugs (e.g., cocaine, heroin, amphetamines, opiates) (often referred to as the “gateway” theory). A measure of the level of drug involvement was created to examine the degree or level of alcohol/drug us e among youths participat ing in this study. A categorical variable was created for Ti me 1 and Time 2 to describe the youths’ level of past year drug involvement. This measure involves four categories: (1) none, (2) used only tobacco and/or alcohol, (3) used ma rijuana and perhaps tobacco or alcohol (not other drugs), and (4) used other drugs (coc aine, amphetamines, barbiturates/sedatives, inhalants, hallucinogens, and opiates) and perh aps tobacco, alcohol, or marijuana. For Time 1, 48 percent of the youths reported they used no drugs in the past year, 15 percent reported using only tobacco or alcohol, 28 pe rcent reported using marijuana and not other drugs, and 9 percent reported using other drugs. For Time 2, 64 percent of the youths

PAGE 153

142 reported they used no drugs in the past year 14 percent reported using only tobacco or alcohol, 15 percent reported us ing marijuana and not other dr ugs, and 8 percent reported using other drugs. The developers of the AIW project have derived a scale of drug use problems that takes into account the progressive nature of juvenile substance use and incorporates the alcohol/drug use problem CASI subscales (for further de tail see Dembo et al., under review; Dembo, Wareham, P oythress, Cook, & Schmeidler, in press). This drug problems scale examines past year drug use, drug involvement, and effects of drug use. The drug problems measure was reproduced for the 137 youths completing the AIW follow-up interview in this study. Comparison of the self-reported drug use levels with urine a nd hair drug test analyses provided a conservative assessment of the validity of the drug use measures. Urine and/or hair drug test re sults were available for 113 (83 %) of the youths for Time 1 and 69 (50%) of the youths for Time 2. A crosstabulation compar ing the four selfreported drug involvement categories with positiv e biological assays s uggests that overall self-report drug use was consistent with the drug test findings. Am ong youths in Time 1 who provided both self-reported drug use and biological assays, the following drug positive rates for one or more drugs were f ound: (1) none (12%), (2) used only tobacco and/or alcohol (0%), (3) used marijuana but not other drugs (63%), and (4) used other drugs (25%) (Fisher’s Exact Test = 21.07, p < .001). Among youths in Time 2 who provided both self-reported drug use and biological assa ys, the following drug positive rates for one or more drugs were found: (1 ) none (0%), (2) used only tobacco and/or alcohol (8%), (3) used marijuana but not ot her drugs (69%), and (4) used other drugs

PAGE 154

143 (23%) (Fisher’s Exact Test = 27.90, p < .001). Th ese findings suggest that the self-report measures of drug/alcohol use among this sample are fairly valid. The developers of the CASI provide several subscales describing alcohol and other drug use problem behaviors. Each s ubscale is created by taking the arithmetic average of responses to three dichotomous questions about past year substance use consequences. The results are continuous censored measures, with responses ranging from 0 to 1 for each subscale. Four of the five CASI subscales were used for this study at Time 1 and Time 2. For the serious consequences subscale, youths were asked to indicate significant periods of time having “repeated arguments with family, friends,…because of substance use,” experiencing “substance-related legal issues ,” and “ having “accidents or…injuries when using substances” (alpha reliability: Time 1 = .49, Time 2 = .51). For the loss of control subscale, youths were asked to indicate signi ficant periods of time where they “continued use while in…situations that were…danger ous,” wanting “to cut down, stop using,” and taking “substance(s) in larger amounts…than or iginally intended” (alpha reliability: Time 1 = .67, Time 2 = .61). For the narrowing of behavior repertoire subscale, youths were asked to indicate whether or not they had significant periods where they “spent a great deal of time in activities…to obtain, inge st or recover from using,” “attended activities…under the influe nce,” and “consistently used instead of going to…doing thing” (alpha reliabil ity: Time 1 = .43, Time 2 = .61). For the physical dependence subscale, participants were asked to report wh ether or not they “had to do more…to feel the same effect,” “experienced withdrawal sy mptoms,” and “used substance(s) to avoid withdrawal” (alpha reliability : Time 1 = .46, Time 2 = .52).

PAGE 155

144 Drug problems. For Time 1 and Time 2, a CFA was conducted on the categorical measure of past year drug usage (created abov e), number of drugs tested positive, and the four CASI drug use and consequences subscal es. Since the drug involvement and drug test positive measures were categorical and there was a re spectable amount of missing data on the drugs assay measure (Time 1: 27%, Time 2: 50%), the factor analyses were produced using Mplus (Muthn & Muthn, 2004) Table 6 reports the CFA results for the drug use problems measure. A confirmatory factor anal ysis, specifying one factor, was completed for Time 1 and Time 2 drug problem measures. Initial CF A results suggested th e fit of both models could be improved (Time 1: chi-square = 24.54, df = 5, p = 0.00; CFI = 0.760; TLI = 0.856, RMSEA = 0.169; Time 2: chi-square = 41.79, df = 4, p = 0.00; CFI = 0.837; TLI = 0.837; RMSEA = 0.263). Modification indices fo r Time 1 suggested that correlations between (a) past year drug usage and drug test results and (b) seri ous consequences and physical dependency problems would improve th e fit of the model. Modification indices for Time 2 also suggested that a correlati on between past year drug use and drug test results would improve the fit of the model. The revised models were found to fit the data rather well (Time 1: chi-square = 5.95, df = 5, p = 0.31; CFI = 0.988; TLI = 0.993 RMSEA = 0.037; Time 2: chi-square = 1.30, df = 4, p = 0.86; CFI = 1.000; TLI = 1.012; RMSEA = 0.000). Each of the variables load ed significantly on the revised factors. Summary factor scores were saved in Mplus for use in testing the hypothesized model (Figure 3). Higher scores on the factors indicated higher dr ug involvement, effects, and negative consequences (alpha reliabi lity: Time 1 = .77, Time 2 = .83).

PAGE 156

145 Table 6: CFA Standardized Loadings for Drug Use Problems for Time 1 and Time 2 Time 1 Time 2 Latent Variable Peer Items Standardized Loadings Standardized Loadings Drug Past year drug use .64 .70 Problems Drug test results .50 .22 Serious consequences .60 .84 Narrowing of behavior repertoire .74 .85 Loss of control .75 .82 Physical dependence .79 .90 Eigenvalue = 2.73 3.44 Variance = 45.5 57.3 Time 1: 2 = 5.95, df = 5, p = 0.31; CFI = 0.988; TLI = 0.993 RMSEA = 0.037. Time 2: 2 = 1.30, df = 4, p = 0.86; CFI = 1.000; TLI = 1.012; RMSEA = 0.000.

PAGE 157

146 Table 7: Descriptive Statistics for Observed Measures Time 1: Baseline Interview Variables Minimum Maximum Mean Standard Deviation Family Disruption -0.6684 2.1623 0.1020 0.6460 Family Abuse -0.3134 1.6886 0.1071 0.4585 Peer Strain -0.5680 2.1619 0.1011 0.6187 Parent Attachment -0.4558 1.1831 0.0492 0.4069 School Attachment -0.5295 1.7630 0.0931 0.6018 School Commitment -0.5503 1.3772 0.0882 0.5286 Delinquent Peers -0.7064 1.7284 0.0793 0.6960 Total Delinquency 0 400 12.0949 42.9827 Total Delinquency (log) -1.00 2.60 0.2741 0.8340 Drug Problems -0.2872 3.8867 0.0067 0.6009 Internalizing 0 13 1.9708 2.8516 Externalizing 0 10 2.5985 2.6050 APSD Impulsivity 0 8 3.58 1.8930 YPI Impulsivity-Irresponsibility 1 39 17.94 8.4330 YPI Impulsivity 0 14 6.42 3.4520 YPI Irresponsibility 0 13 3.86 3.2790 YPI Thrill-Seeking 1 15 7.66 3.4900 (Continued on the next page)

PAGE 158

147 Table 7: (Continued) Time 2: Follow-Up Interview Variables Minimum Maximum Mean Standard Deviation Family Disruption -0.1815 1.8851 0.1264 0.4849 Family Abuse -0.0834 0.9222 0.0395 0.2368 Peer Strain -0.4399 1.8603 0.1313 0.5732 Parent Attachment -0.3297 1.0089 0.0563 0.3761 School Attachment -0.4951 1.7965 0.1247 0.6580 School Commitment -0.4163 1.1864 0.0652 0.4670 Delinquent Peers -0.5556 1.5080 0.0943 0.6243 Total Delinquency 0 911 22.6350 101.7792 Total Delinquency (log) -1.00 2.96 -0.2789 1.0195 Drug Problems -0.2462 3.9414 0.0067 0.6764 Internalizing 0 7 1.1314 1.7185 Externalizing 0 5 1.2482 1.3974 APSD Impulsivity --------YPI Impulsivity-Irresponsi bility --------YPI Impulsivity --------YPI Irresponsibility --------YPI Thrill-Seeking --------Description of Observed Variables The mean, minimum, maximum, and standard deviation values for the Time 1 and Time 2 strain, social control, differential association, delinquency (both before and after log transformation), drug problems, and persona lity measures are reported above in Table

PAGE 159

148 7. In general, the measures reflect factor sc ores. Therefore, the scale for these measures is relatively small, based on the range of possi ble values. On average, the youths in this sample reported low levels of strain, social control, delinquent peers, internalizing behavior, and externalizi ng behavior. The youths reported average impulsivity characteristics that were modest in size. Bi variate correlations for these final measures are shown in Appendix B. Most of the meas ures were significantly correlated in the expected direction. The next chapter describes the analytic strategy employed in the present study. As mentioned previously, the small size of the sample used in this study limited the complexity of the analyses conducted herein. In most cases, th e measures described above were used to create latent measures. Since the “observed” measures were factor scores created through CFA, the latent measur es may be conceptualized as second-order factor analyses. Findings from the analyses are reported and interpreted in the next chapter.

PAGE 160

149 Chapter 6 Results Chapter 5 discussed the steps taken to cr eate the Time 1 and Time 2 measures of strain, social control, differential a ssociation, delinquency, drug problems, and personality characteristics. Since the samp le used in this study is limited to only 137 cases for Time 1 and Time 2, any analyses of this data must be parsimonious. Figures 2 and 3 (described in Chapter 4) reflect an attempt to maintain parsimony, while maximizing the empirical and theoretical c ontribution of this study. Unlike Agnew et al.’s (2002) study of strain and personality, this study is limited to examining key latent and observed variables, without controlling for influential demographic characteristics (e.g., age, race/ethnicity, gender, SES). Analytic Strategy The present study tests longitudinal st ructural equation models (SEM) (see Figures 2 and 3 in Chapter 4) of strain on delinquency and drug problem behaviors. Structural equation modeling is a statistical technique used to examine a priori specified relationships between both observed and unobserve d (i.e., latent) measur es (for detail see Bollen, 1989; Byrne, 2001). SEM provides the opportunity for gr aphic or pictorial description of the relatio nship between variables that repres ent a series of structural (i.e., regression) equations. Gene rally, the SEM model can be described as containing two submodels: a structural model and a measuremen t model. The structural model describes

PAGE 161

150 the relationships between the latent variables (though con ceptually, observed variables may also be treated as latent and included in the structural model). The measurement model describes the relationship between the observed and latent variables. That is, it describes how the observed measures load onto or relate to the latent variables. The measurement model was described in Chapter 4. Due to the complexity of the models in Figures 2 and 3, the measurement models are not included in the SEM illustration. The SEM models show latent variables for strain, social control, delinquent peers, a nd drug problems. The structural model for Time 1 and Time 2 strain contains three “obser ved” factors of strain estimated as loading onto one overall latent measure. Family disr uption, family abuse, and poor peer relations factor scores are estimated as the latent variable, strain The structural model for the latent variable, social contro l, is hypothesized to be co mprised of the following three “observed” factor scores : parental attachment, school attachment, and school commitment. Delinquent peers and drug problem s were included as “observed” factor scores, rather than latent meas ures. This data reduction to second-order latent variables helped to lower the number of parameter estimates required to successfully compute the full SEM models. The SEM analyses were completed usi ng Mplus version 3.12 (Muthn & Muthn, 2004). Mplus is a versatile, sophisticated st atistical modeling program. It permits the estimation of continuous and categorical, both observed and latent variables. Mplus provides a chi-square test of the null hypothesis to test the fit of m odels to their data. Lack of significance for the chi-square indi cates an acceptable fit of the model to the data. The software also provides a number of goodness of fit m easures to assess the

PAGE 162

151 closeness of the fit of the m odel to the data (e.g., CFI, TLI, RMSEA). Mplus also allows the application of various estimators (e .g., maximum likelihood estimation) of the parameters in the model, some of which provid e chi-square test statistics that are robust to non-normality in the data. The premise of this study is to exam ine the relationship between strain and maladaptive personality characteristics on de linquency and drug use problems within a GST framework. As such, a stepwise appr oach was taken, beginning with the most simplistic GST model and progressing to ri cher, more complex models—eventually to those illustrated in Figures 2 and 3. Assumi ng the relationship between the latent strain measure and delinquency/drug use problems were statistically significant, the social control, differential associat ion, and personality measures were then examined in subsequent models. Findings Initial GST Models The first step was to address the hypothesi s that strain at Ti me 1 has significant positive effects on delinquency/drug problems at Time 2. Preliminary analyses were conducted testing the effects of strain on the self-reporte d total delinquency index (log transformed) and drug use problems factor. First simple models were estimated for the strain measures on delinquency and dr ug problems, respectively. Then more complex models examining the influence of social cont rol and delinquent peer association factors on the strain-delinquency and strain drug problems relations hips were examined. As Tables 8 and 9 show, the analyses indi cated that three of the four models did not fit the data. Indeed, the delinquency m odels could not be estimated properly due to

PAGE 163

152 negative residuals caused by latent correlation values greater than or equal to 1, which suggests linear dependency, and data conve rgence issues. For the drug use problems models, the simple model was estimated normally after correlating the error terms between poor peer relations items for the latent strain variables across Time 1 and Time 2 (unstandardized estimate = 0.125, standard ized estimate = 0.354, p < .05). Although the simple model for the drug problems factor fit the data well after making this modification (Chi-square = 20.30, df = 16, p = .21, CFI = 0.979, TLI = 0.962, RMSEA = 0.044), the relationship between strain at Time 1 and drug use problems at Time 2 did not attain statistical significance (using a one-tailed test). Drug problems at Time 1 were significantly and positively rela ted to drug problems at Time 2. Strain at Time 1 was significantly and positively related to strain at Time 2. The model explained 25 percent of the variance in drug problems at Time 2 (R2 = 0.251). None of the complex models in Tables 8 and 9 could be estimated due to data convergence issues.

PAGE 164

153 Table 8: Delinquency (log) on Strain, Soci al Control, and Delinquent Peer Factor s Estimates (Standardized Estimates) Simple Model Complex Model Delinquency (Log) Delinquency (Log) Endogenous Variables Strain (T1) Strain (T2) Delinquency (T1) Strain (T1) Strain (T2) Social Control (T1) Social Control (T2) Delinquent Peer (T1) Delinquency (T1) Family Disruption 1.000 (0.846) 1.000 (0.527) Family Abuse 0.520** (0.620) 0.178* (0.192) Peer Strain 0.451** (0.399) 0.797** (0.355) (No convergence. Number of iterations exceeded.) Parental Attachment School Commitment School Attachment Strain (T2) 0.485** (1.038)a Social Control (T2) Delinquent Peer (T2) Delinquency (Log) (T2) 0.114 (0.062) 0.420** (0.348) Note. p < .10, ** p < .05. a. Model could not be fully estimated. Correlation exceeds 1.

PAGE 165

154 Table 9: Drug Problems Factor on Strain, Social Control, and Delinque nt Peer Factors Estimates (Standardized Estimates) Simple Model Complex Model Drug Problems Factor Drug Problems Factor Endogenous Variables Strain (T1) Strain (T2) Drug Problems (T1) R2 (T1) R2 (T2) Strain (T1) Strain (T2) Social Control (T1) Social Control (T2) Delinquent Peer (T1) Drug Problems (T1) Family Disruption 1.000 (0.938) 1.000 (0.558) 0.879 0.311 Family Abuse 0.456** (0.602) 0.111 (0.127) 0.363 0.016 Peer Strain 0.352** (0.345) 0.718** (0.338) 0.119 0.114 (No convergence. Number of iterations exceeded.) Parental Attachment School Commitment School Attachment Strain (T2) 0.407** (0.912) Social Control (T2) Delinquent Peer (T2) Drug Problems (T2) 0.186* (0.167) 0.469** (0.419) 0.251 Note. p < .10, ** p < .05. Simple Model: Chi-square = 20.30, df = 16, p = .21, CFI = 0.979, TLI = 0.962, RMSEA = 0.044.

PAGE 166

155 SEM of GST for Time 1 Only The simple models described in the above tabl es suggested that Time 1 and Time 2 strain were linearly dependent (i.e., latent co rrelation greater than or equal to 1). (The strain T2 on strain T1 standard estimate was 1.038 for the delinquency model and 0.912 for the drug problems model.) Negative vari ance or residual variance can sometimes occur when models are not properly specified (Muthn & Muthn, 2004). Therefore, the analyses proceeded by examining the models described in Figures 2 and 3 without the Time 2 strain, social control, and de linquent peer association measures. Table 10 describes the findings for the Ti me 1 strain only (simple) models. The model testing self-repo rted total delinquency (log transf ormed) regressed on strain at Time 1 fit the data moderately well (C hi-square = 7.44, df = 4, p = .11, CFI = 0.967, TLI = 0.917, RMSEA = 0.079). The observed strain ite ms loaded significantly and positively on the latent strain variable. Strain at Time 1 did not significantly predict the measure for self-reported delinquency (estimate = 0.017, critical-ration = 0.120). Delinquency at Time 1 had a significant positive effect on delinquency at Time 2 (estimate = 0.489, critical-ratio = 4.827). The model explained only 16 percent of the variance in Time 2 delinquency (R2 = 0.163). The model examining the effects of Time 1 strain on drug use problems could not be properly estimated due to extreme latent correlations (see loading/standardized estimate for family disruption item on stra in T1: standardized estimate = 1.040). Since the Time 1 simple models reflect a minimalis t approach, it is unlikely that the model suffers from further misspecification. Mplu s can be sensitive to the distribution of variable values; therefore, it is more likely that the measures themselves are problematic.

PAGE 167

156 Table 10: Delinquency (log) and Drug Problems Factor on Time 1 Strain Estimates (Standardized Estimates) Simple Model Simple Model Delinquency (Log) Drug Problems Factor Endogenous Variables Strain (T1) Delinquency (T1) R2 Strain (T1) Drug Problems (T1) Family Disruption 1.000 (0.972) 0.945 1.000 (1.040)a Family Abuse 0.424** (0.581) 0.337 0.371** (0.544) Peer Strain 0.332** (0.337) 0.114 0.293** (0.318) Delinquency (T2) 0.017 (0.011) 0.489** (0.400) 0.163 Drug Problems (T2) 0.133 (0.132) 0.502** (0.446) Note. p < .10, ** p < .05. a. Model could not be fully estimated. Correlation exceeds 1. Delinquency: Chi-square = 7.44, df = 4, p = .11, CFI = 0.967, TLI = 0.917, RMSEA = 0.079. Variable Adjustment In an effort to minimize any estimations problems that are a consequence of the variance and distribution of the measures, the certain measures were adjusted or recoded to improve the estimation of the models. The variable adjustment process began by recoding and adjusting the delinquency and dr ug use measures. Three new measures for delinquency and one measure for drug use were created. The self-reported total delinquency log transformed measure was firs t altered by shifting the distribution one unit to the right of the y-axis. This wa s accomplished by increasing the log transformed value by one. The result was a delinquency m easure that started at zero, but maintained the same conceptual agreement regarding rates of offending as the original log

PAGE 168

157 transformed measure. Next, delinquency was recoded into four categories: 0 (delinquency = 0), 1 (delinquency = 1 to 10), 2 (delinquency = 11 to 100), and 3 (delinquency = 101 to 1000). Finally, de linquency was recoded as a dichotomous measure (0 = 0 offenses, 1 = 1 or more o ffenses). The past year drug use measure containing four categories (none, used only t obacco and/or alcohol, used marijuana and perhaps tobacco or alcohol, and used ot her drugs and perhaps tobacco, alcohol, or marijuana) was used in place of the drug problem factor (see Chapter 5 for description). As mentioned earlier, this scale is consistent with the “gate way” drug literature (Kandel, 1975, 2002; Kandel et al., 1978; Kandel et al., 19 92). Table 11 provides the descriptive statistics for these new measures.

PAGE 169

158 Table 11: Descriptive Sta tistics for Adjusted Delinque ncy and Drug Measures Time 1(Baseline) Variables Minimum Maximum Mean Standard Deviation Delinquency (log + 1) 0 3.6000 1.2741 0.8340 Delinquency (categorical) 0 3 1.0100 0.6910 Delinquency (0/1) 0 1 0.8000 0.4050 Drug Use Level 0 3 0.9700 1.0570 Time 2(Follow-Up) Variables Minimum Maximum Mean Standard Deviation Delinquency (log + 1) 0 3.9600 0.7211 1.0195 Delinquency (categorical) 0 3 0.5900 0.8540 Delinquency (0/1) 0 1 0.4100 0.4930 Drug Use Level 0 3 0.6700 1.0010 Despite the recodes for the self-reporte d delinquency and substance use measures, SEM analyses for these measures regresse d on strain (Time 1 and Time 2) were unsuccessful. Each model continued to expe rience issues with the Time 1 and Time 2 strain relationships. The estimates for the models are presented in Table 12. Once again, the models could not be estimated due to negative variance or correlations exceeding scores of 1 between strain Time 1 and strain Time 2: for delinquency (log + 1) r = 1.038; for delinquency (categorical), r = 1.142; for delinquency (dummy), r = 1.051; and for drug use (categorical), r = 1.118. Moreover, in the adjusted delinquenc y models strain at Time 1 continued to be non-significantly re lated to delinquency at Time 2. The revised

PAGE 170

159 drug use measure, however, did appear to improve the estimation of Time 2 drug use level on strain at Time 1. The models were replicated using Time 1 strain only, and the results were similar to those reported in Table 12 for the three de linquency measures. As Table 13 shows, the models for the adjusted delinquency meas ures for categorical and dichotomous classification and drug use cate gories could not be properly es timated due to correlations exceeding values of 1 on family disruption. The results for the log shift measure of delinquency were the same as the log tr ansformed delinquency model in Table 10—as expected. Strain was not a significant predictor of 12-month follow-up delinquency. The variable adjustments made to th e delinquency and drug problems measures were intended to improve the fit of the full Time 1 and Time 2 strain model to the data (i.e., improve estimation). Unfortunately, operational changes did not improve the estimations of the models. The models remained too complex for the data.

PAGE 171

160 Table 12: Recoded Delinquency and Drug Use on Strain Estimates (Standardized Estimates) Delinquency (Log+1) Delinquency (Categorical) Endogenous Variables Strain (T1) Strain (T2) Delinquency (T1) Strain (T1) Strain (T2) Delinquency (T1) Family Disruption 1.000 (0.849) 1.000 (0.527) 1.000 (0.783) 1.000 (0.528) Family Abuse 0.520** (0.620) 0.178* (0.192) 0.509** (0.562) 0.163* (0.177) Peer Strain 0.451** (0.399) 0.797** (0.355) 0.512** (0.418) 0.794** (0.355) Strain (T2) 0.485** (1.038)a 0.578** (1.142)a Delinquency (T2) 0.144 (0.619) 0.420** (0.348) 0.162 (0.607) 0.536** (0.369) Delinquency (Dummy) Drug Use (Categorical) Endogenous Variables Strain (T1) Strain (T2) Delinquency (T1) Strain (T1) Strain (T2) Drug Use (T1) Family Disruption 1.000 (0.834) 1.000 (0.525) 1.000 (0.828) 1.000 (0.484) Family Abuse 0.506** (0.594) 0.148 (0.159) 0.470** (0.548) 0.134 (0.132) Peer Strain 0.465** (0.405) 0.817** (0.363) 0.502** (0.434) 0.967** (0.396) Strain (T2) 0.497** (1.051)a 0.491** (1.118)a Delinquency (T2) 0.376 (0.202) b Drug Use (T2) 0.377** (0.201) 0.527** (0.555) Note. p < .10, ** p < .05. a. Model could not be fully estimated. Correlation exceeds 1. b. Could not be calculate d. Caused a singular we ight matrix error.

PAGE 172

161 Table 13: Recoded Delinquency and Drug Us e on Strain (Time 1 Only) Estimates (Standardized Estimates) Delinquency (Log+1) Delinquency (Categorical) Endogenous Variables Strain (T1) Delinquency (T1) R2 Strain (T1) Delinquency (T1) Family Disruption 1.000 (0.972) 0.945 1.000 (1.137)a Family Abuse 0.424** (0.581) 0.337 0.309 (0.480) Peer Strain 0.332** (0.337) 0.114 0.207 (0.240) Delinquency (T2) 0.017 (0.011) 0.489** (0.400) 0.163 0.082 (0.053) 0.641** (0.405) Delinquency (Dummy) Drug Use (Categorical) Endogenous Variables Strain (T1) Delinquency (T1) Strain (T1) Drug Use (T1) Family Disruption 1.000 (1.041)a 1.000 (1.113)a Family Abuse 0.370** (0.538) 0.317** (0.484) Peer Strain 0.248* (0.277) 0.246* (0.277) Delinquency (T2) 0.214 (0.140) 0.416 (0.166) Drug Use (T2) 0.279 (0.150) 0.753** (0.623) Note. p < .10, ** p < .05. a. Model could not be fully estimated. Correlation exceeds 1. Delinquency (Log+1): Chi-square = 7.44, df = 4, p = 0.11, CFI = 0.967, TLI = 0.917, RMSEA = 0.079.

PAGE 173

162 In a final effort to improve the fit of the SEM models, the family, peer, and school observed measures were recoded. Similar to Wallace et al.’s (2005) recent GST test, additive indexes of strain measures were cr eated using the EFA and CFA factor loadings as a measurement guide. This transformati on had no effect on the internal consistency (i.e., alpha reliability) of the strain (family disruption, family abuse, and poor peer relations), social control (low parental attachment/commitment, low school attachment, and low school commitment), and delinquent pe er association indexes. However, it did change the scale of the measures, there by making them less problematic for conducting the SEM analyses. These scores are integers, rather than small continuous factor scores, reflecting a summary of the CASI family, p eer, and school life area items described in Chapter 5. Table 14 provides th e descriptive statistics for th e strain, social control, and differential association indexes. Table 14: Descriptive Statistics for Strain, So cial Control, and Soci al Learning Indexes Time 1(Baseline) Variables Minimum Maximum Mean Standard Deviation Strain: Family disruption 0 13 1.68 2.521 Family abuse 0 6 0.45 1.063 Peer strain 0 13 1.85 2.600 Social Control: Parental attachment 0 15 3.15 3.420 School attachment 0 8 1.93 2.010 School commitment 0 6 1.13 1.571 Social Learning: Delinquent peers 0 8 2.18 2.355 (Continued on the next page)

PAGE 174

163 Table 14: (Continued) Time 2(Follow-Up) Variables Minimum Maximum Mean Standard Deviation Strain: Family disruption 0 7 0.71 1.219 Family abuse 0 2 0.21 0.430 Peer strain 0 7 1.15 1.652 Social Control: Parental attachment 0 10 1.82 2.104 School attachment 0 4 1.18 1.253 School commitment 0 2 0.28 0.610 Social Learning: Delinquent peers 0 4 1.20 1.367 The Time 1 and Time 2 strain simplifi ed model proved problematic, despite all efforts to recode the measures. Analyses consistently revealed linear dependency problems (high multicollinearity) for the strain measures over time. These results remained consistent even when estimators robust to non-normality in the data (e.g., weighted least square mean variance [W LSMV] and maximum likelihood with robust errors [MLR]) were used (see Muthn & Muthn, 2004). The robust estimators were thought to be an acceptable application gi ven the slight skewness of the family disruption, family abuse, and parental at tachment measures. Consequently, all subsequent analyses were conducted using Time 1 endogenous measures and Time 2 delinquency and drug use/problems measures. As seen in Table 15, the simplified SEM m odels predicting that strain at Time 2, as measured by the family, peer, and school ad ditive indexes, lead s to delinquency (log transformed) and drug use fit the data quite well (delinquency: chi-square = 3.04, df = 4,

PAGE 175

164 p = .55, CFI = 1.000, TLI = 1.060, RMSEA = 0.000; drug use: chi-square = 0.50, df = 3, p = .92, CFI = 1.000, TLI = 1.155, RMSEA = 0.000). (The drug problems model is not reported in Table 15 because it experienced negative residuals between Time 1 and Time 2 drug use.) Delinquency at Time 1 significantly predicted higher delinquency at Time 2. Drug use at Time 1 significantly predicted higher drug use leve ls (i.e., progressive use of drugs) at Time 2. For the delinquency and dr ug use models, strain at Time 1 does not, however, significantly predict delinquency or drug use in the follow-up year. Even more noteworthy was the fact that th e three strain index measures do not load significantly onto the strain latent variable. The only strain measure that loaded si gnificantly on the strain at Time 1 was poor peer relations for the de linquency model. This underscores the lack of cohesion in the latent variab le strain. Obviously, the vari ables used to measure strain at Time 1 and Time 2 (both factor scores a nd indexes) lack sufficient conceptual cohesion to be estimated by one overall latent measure. After extensive data manipulation, the SEM analyses testing the effects of strain at Time 1 on delinquency/drug use at Time 2 failed to support hypothesis #1. For the justice referred youths in this sample, st rain does not directly affect self-reported delinquency or drug use problems. As a re sult, hypotheses 2 through 6 were also not supported by the SEM models.

PAGE 176

165 Table 15: Delinquency (log) and Drug Problems Factor on Time 1 Strain Index Estimates (Standardized Estimates) Simple Model Simple Model Delinquency (log) Drug Use Factor Endogenous Variables Strain (T1) Delinquency (T1) R2 Strain (T1) Drug Use. (T1) R2 Family Disruption 1.000 (0.586) 0.343 1.000 (0.854) 0.730 Family Abuse 0.113 (0.157) 0.025 0.073 (0.148) 0.022 Peer Strain 0.783** (0.445) 0.198 0.368 (0.305) 0.093 Delinquency (T2) 0.019 (0.027) 0.479** (0.392) Drug Use (T2) 0.093 (0.200) 0.548** (0.577) Delinquency (T1) 0.163 Drug Use (T1) 0.426 Note. p < .10, ** p < .05. Delinquency: Chi-square = 3.04, df = 4, p = .55, CFI = 1.000, TLI = 1.060, RMSEA = 0.000. Drug Use: Chi-square = 0.50, df = 3, p = .92, CFI = 1.000, TLI = 1.155, RMSEA = 0.000. Supplemental Analyses Path Analyses of Strain Leading to Delinquency/Drugs As the preceding section illustrates, th e SEM analyses of the GST models in Figures 2 and 3 were not sufficiently specified. In part, this difficu lty was caused by the limited sample size (n=137), which complicated parameter estimation. The small sample size meant that the hypothesized models also had to be limited in their scope. Another limitation contributing to the SEM outcomes wa s the types of measures available to operationalize the strain constructs, irrespec tive of social control and differential

PAGE 177

166 association measures. Given these obstacl es and the unaccommodating SEM results, one option was to conclude that hypotheses 2 th rough 6 were summarily unanswerable using these data. An alternative option was to re-s pecify the models. The latter approach was taken. The models were re-specified and supplemental analyses of GST and personality were pursued employing path analyses of th e individual indexes comprising the latent strain and social control va riables, as well as the deli nquent peer association index. Utilizing the same “bottoms-up” approach employed in the SEM analyses, simple models of past year self-reported delinque ncy (log transformed) (DELINQ) and drug problems (DRUGPROB) or drug usage (DRUGU SE) at Time 2 were regressed on the three Time 1 strain indexes: family di sruption (FAMDIS), family abuse/neglect (FAMABUS), and poor peer relationships (POORPEER). Mplus version 3.12 (Muthn & Muthn, 2004) was used to conduct the path analyses of GST and personality characteristics. Table 16 reports the findings for the ba sic path analysis models regressing delinquency and drug problems/use on the three indexes of strain. All three models found no significant relationships between stra in at Time 1 and deviance at Time 2. Since the models are just-identified (i.e., all parameters specified in the model), the chisquare test and goodness of fit measures could not be calculated for the basic strain path models. Yet the size and non-significance of the estimates corroborate the SEM findings.

PAGE 178

167 Table 16: Unstandardized Estimates for Path Analyses of Se lf-Reported Delinquency and Drug Problems-Usage on Strain (T1) Exogenous Variables Endogenous Variables Family Disruption Family Abuse Peer Strain Delinquency Drug Problems Drug Use R2 Model 1: Delinquency (log) (T2) on strain (T1) & delinquency (T1) Delinquency -0.007 0.100 0.004 0.490** 0.173 Model 2: Drug problems factor (T2) on strain (T1) & delinquency (T1) Drug problems 0.035 0.055 0.004 0.497** 0.263 Model 3: Drug use level (T2) on strain (T1) & delinquency (T1) Drug use 0.086* -0.062 0.021 0.745** 0.428 Note. p < .10, ** p < .05.

PAGE 179

168 Path Analyses of Delinque ncy/Drugs Leading to Strain Scholars have suggested that criminologica l theories can be organized according to the emphasis they place on individual di fferences (Johnson, Hoffmann, Su, & Gerstein, 1997). According to Johnson et al. (1997), th eories will utilize either a “population heterogeneity” or “state dependence” (Nag in & Farrington, 1992; Na gin & Paternoster, 1991) approach when examining indivi dual differences. From a “population heterogeneity” perspective, individual di fferences in crime/delinquency result from developmental differences that arise early in life (i.e., childhood and adolescence) and remain somewhat stable over the life-course From a “state dependence” perspective, individual differences in crime/delinquen cy result from changes that arise as a consequence of committing a criminal/deviant act (e.g., changes in perceptions of costs and benefits of crime). In general, GST advocates population hetero geneity as the etio logical explanation for deviance. Individuals are driven to commit their first and subsequent deviant acts by the need to quickly and easily relieve the pressure of unsatisfied or blocked desires. Differences in the magnitude, duration, a nd frequency of strain, conditioned by developmental differences in coping mechan isms and other conditioning factors (e.g., personality), result in different trajectories of crime. Accordingly, differences in strainful char acteristics, as well as social control, differential association, and pers onality characteristics, should have significant effects on delinquency and drug problems/use among youths in this study. Almost all of the participants were first-time official offenders (first documented arrest or charge) (91%) prior to their Arbitration dive rsion program offense. Almost all of the cases (90%) had

PAGE 180

169 baseline interviews completed within 60 da ys of the program offense. Since the questions pertaining to strain referred to events and conditi ons that occurred 12 months prior to the initial interview, it is likely that the strainful events capture characteristics that are precursors, rather than consequences, of delinquent behavior. Therefore, a population heterogeneity explanation of cr ime should have predicted delinquency and drug use at Time 2 based on Time 1 strain. One of the potential explanations for the nul l findings of the baseline strain effects on follow-up year delinquency and drug problem s/use may lie in the nature of the sample. As mentioned in the description of the demographic characteristics of this sample, many of these youths come from relatively low to modest SES backgrounds, and have experienced prolonged peri ods of school and family diffic ulties. Therefore, it may be less the case that the baseline intervie w strain reflects problematic conditions, and more the case that these experiences have been somewhat prolonged, almost normal circumstances. Assuming this is true, populat ion heterogeneity for this sample may not provide a viable explanation for delinquency and drug use. Among this sample of youths, a better explanation for criminal propensity may require a state dependence approach to GST. A state dependence explanation of GST would hypothesize that delinquenc y results in negative consequences that increase the likelihood of subsequent strain and delinquenc y. That is, delinquency causes both strain and future delinquency. Youths who report higher levels of delinquency/drug use at Time 1 should be more likely to experience subsequent strain and delinquency/drug use than those reporting little or no delinquency/drug use.

PAGE 181

170 In the GST literature, two studies, in particular, have examined the longitudinal, reciprocal effects of stressful life experi ences on subsequent experiences of strain (Aseltine et al., 2000; Kim et al., 2003). In both of these studies, strain and delinquency were considered to be both predictors and outcomes of previous and subsequent strain and delinquency. Overall, thes e studies reported that early experiences of strain led to delinquency, which, in turn, led to higher levels of strain. Further, other research has demonstrated that delinquent activities can lead to increased stress and strain (e.g., Elliott et al., 1985, 1989; Herrenkohl et al., 2000; Sampson & Laub, 1993, 2003). Based on the body of literatur e that suggests that deli nquency may cause and/or exacerbate subsequent strain, the models examined in the present study were modified for ad hoc analyses. In the ad hoc model, Time 1 delinquency and drug use (not shown in Figure 4) were hypothesized to lead to de linquency/drug use and strain at Time 2. Personality characteristics at Time 1 were hypothesized to significantl y affect strain at Time 2. Since GST postulates that the nexus between strain and delinquency may be rather contemporaneous (Agnew, 1992), strain at Time 2 was predicted to significantly and positively affect delinquency/drug use at Time 2. Strain at Time 2 was hypothesized to be a partial mediator of the relationship between pers onality characteristics and delinquency. Similar to Figures 2 and 3, social control and differentia l association/social learning theory are viewed as complementary theories. Figure 4 provides an interpretation of this ad hoc model for delinquency. The same model will be used to examine the effects of strain on drug problems and drug use (replacing the term delinquency with the appropriate drug m easure). Moderating effects of personality characteristics at Time 1 on the strain-deviance Ti me 2 relationship will

PAGE 182

171 also be examined by including interaction measures in the model. Any significant findings from this model regarding stra in and delinquency/drug problems should be considered cross-sectional in nature. Theref ore, the causal nature of subsequent findings must be viewed and interpreted with caution.

PAGE 183

172 Figure 4: Ad Hoc Contemporaneous Model of Strain, Social Cont rol, and Delinquent Peers Path analyses for self-reported delinqu ency and drug problems/use were carried out in a stepwise manner. Mplus versi on 3.12 (Muthn & Muthn, 2004) was used to perform the analyses, which included applica tion of the missing imputation feature. As seen in Table 17, a total of 23 models were examined for delinquency. All of the models have marginal goodness-of-fit values in term s of the chi-square, CFI, TLI, and RMSEA values (see bottom of Table 17). In the first step, self-reported delinquency at Time 1 and the three measures of strain at Time 2 we re regressed on self-reported delinquency at Time 2 (see Model 1). Next, self-reported de linquency at Time 1 and the three measures DelinquencyT1 DelinquencyT2 PersonalityT1 Family DisruptionT2 Parental AttachmentT2 Delinquent PeersT2 School AttachmentT2 School CommitmentT2 Peer StrainT2 Family AbuseT2 Personality X Strain InteractionsT1

PAGE 184

173 of social control and delinquent peers at Time 2 were regressed on self-reported delinquency at Time 2 (see Model 2). (School attachment was excluded from the model. Bivariate correlations indicated school attachment and school commitment suffered from multicolinearity issues [r = .919].) The soci al control and delinquent peer association measures explained twice as much of the vari ance in delinquency (39%) at Time 2 as the simple strain model (R2 = 0.16). Simple models regressing delinquency at Time 2 on the personali ty characteristics and psychopathy impulsivity dimensions are reported in Models 3 through 8. The internalizing and externalizing behavior measur es for Time 1 did not significantly relate to delinquency at Time 2 (see Model 3). Th e psychopathy measures for impulsivity were all significantly and positively related to delinquency in the following year (see Models 4 to 8). The personality and psychopathy ch aracteristics explained approximately 20 percent of the variance in delinquency at Time 2. Secondly, the effects of delinquency leadi ng to strain, leadin g to delinquency were examined (see Model 9). Delinquency at Time 1 was significantly related to family disruption at Time 2. Family abuse at Ti me 2 was significantly related to Time 2 delinquency. The addition of the strain m easures predicting Time 2 delinquency nearly doubled the explained variance (R2 = 0.28) of the model for delinquency at Time 2. Time 2 effects for the social control and delinque ncy measures were similarly tested (see Model 10). Higher delinquency at Time 1 significantly predicted low parental attachment, low school commitment, and higher delinquent peer associations at Time 2, which were significantly related to Time 2 delinquency.

PAGE 185

174 In Model 11, the strain items are combined with the social co ntrol and differential association measures. The relationships desc ribed in Models 9 and 10 remain significant and relatively the same in magnitude. This combined model explained 46 percent of the variance in delinquency at Time 2. Models 12 through 17 report the effects of personality character istics at Time 1 on Time 2 strain measures. Internalizing beha viors was significantly and positively related to family disruption. The YPI impulsivity-i rresponsibility and irre sponsibility indexes were significantly and positively related to poor peer relations. However, poor peer relations was not significantly related to Time 2 delinquency. Moreover, the addition of the personality measures did not substantially improve the explained variance of the models. Finally, the psychopathy/personality m easures were included in the model examining the effects of strain, social cont rol, and delinquency as mediators for Time 1 and Time 2 delinquency (see Models 18 th rough 23). Internalizing behaviors was significantly and positively related to family disruption. The YPI impulsivityirresponsibility and irresponsib ility indexes were significantl y and positively related to poor peer relations. However, poor peer relati ons was not significantly related to Time 2 delinquency. The YPI impulsivity-irrespons ibility domain was also significantly and positively related to family disruption. The addition of the personality measures did not substantially improve the explai ned variance of the models. The indirect effects of Time 1 psycho pathy and personality on delinquency at Time 2 via strain were examined using Mplus. Mplus is capable of reporting the partial and total indirect effects of regression path ways. None of the indirect estimates were

PAGE 186

175 significant for the models presented in Tabl e 17. Therefore, the specific psychopathic and personality characteristics examined in this study do not significantly mediate the strain-delinquency relationship. In addition to the mediating role of pers onality characteristic s, the ad hoc path analyses attempted to examine the moderati ng influence of personality on strain. Interaction terms were created by multip lying the psychopathy, internalizing, and externalizing scores by the strain indexes. Then these interaction terms were included in the full models of delinquency. Unfortunately, none of the interaction models fit the data well ( internalizing/externalizing: chi-square = 165.60, df = 45, p = 0.00, CFI = 0.702, TLI = 0.444, RMSEA = 0.140; APSD impulsivity: chi-square = 131.46, df = 25, p = 0.00, CFI = 0.709, TLI = 0.349, RMSEA = 0.176; YPI impulsivity-irresponsibility: chi-square = 129.03, df = 24, p = 0.00, CFI = 0.713, TLI = 0.331, RMSEA = 0.179; YPI impulsivity: chi-square = 106.89, df = 56, p = 0.00, CFI = 0.754, TLI = 0.425, RMSEA = 0.159; YPI irresponsibility: chi-square = 107.69, df = 24, p = 0.00, CFI = 0.759, TLI = 0.438, RMSEA = 0.160; YPI thrill-seeking: chi-square = 138.67, df = 24, p = 0.00, CFI = 0.686, TLI = 0.267, RMSEA = 0.187). Further, the modi fication indices for these models did not provide any theoretically meaningful r ecommendations for improving the fit of the models. The specific psychopathic and persona lity characteristics examined in this study do not significantly moderate the strain-delinquency relationship.

PAGE 187

176 Table 17: Unstandardized Parameter Estimates of the Path Models of Delinquency (log), Strain (T 2), and Personality Characterist ics (T1) (N = 137) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 1: Strain (T2) on delinquency (T1) only Famdis 0.481** 0.108 Famabus 0.076* 0.022 Poorpeer 0.247 0.016 Delinq 0.493** 0.163 Goodness of Fit Measures: Just-identified model. No goodness of fit statistics available. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 188

177 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 2: Social control & delinquent peers (T2) on delinquency (T1) only Paratt 1.064** 0.178 Sklcmt 0.385** 0.064 Delpeer 0.475** 0.084 Delinq 0.493** 0.391 Goodness of Fit Measures: Just-identified model. No goodness of fit statistics available. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 189

178 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 3: Internalizing & externalizing characteristics Delinq 0.464** 0.039 0.261 0.182 Goodness of Fit Measures: Just-identified model. No goodness of fit statistics available. Model 4: APSD impulsivity characteristics Delinq 0.404** 0.093** 0.187 Goodness of Fit Measures: Just-identified model. No goodness of fit statistics available. Model 5: YPI impulsivity-irresponsibility domain Delinq 0.332** 0.033** 0.220 Goodness of Fit Measures: Just-identified model. No goodness of fit statistics available. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 190

179 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 6: YPI impulsivity dimension Delinq 0.405** 0.055** 0.192 Goodness of Fit Measures: Just-identified model. No goodness of fit statistics available. Model 7: YPI irresponsibility dimension Delinq 0.398** 0.067** 0.203 Goodness of Fit Measures: Just-identified model. No goodness of fit statistics available. Model 8: YPI thrill-seeking dimension Delinq 0.377** 0.063** 0.200 Goodness of Fit Measures: Just-identified model. No goodness of fit statistics available. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 191

180 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 9: With delinquency (T2) on strain Famdis 0.477** 0.106 Famabus 0.076* 0.022 Poorpeer 0.247 0.016 Delinq 0.390** 0.061 0.772** 0.063 0.284 Goodness of Fit Measures: Chi-square = 3.60, df = 2, p = 0.16; CFI = 0.975; TLI = 0.876; RMSEA = 0.076. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 192

181 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 10: With delinquency (T2) on social control & delinquent peers Paratt 1.064** 0.182 Sklcmt 0.386** 0.064 Delpeer 0.475** 0.084 Delinq 0.212** 0.112** 0.216** 0.165** 0.391 Goodness of Fit Measures: Chi-square = 4.12, df = 1, p = 0.04; CFI = 0.977; TLI = 0.771; RMSEA = 0.151. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 193

182 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 11: With delinquency (T2) on strain, social control, & delinquent peers Famdis 0.518** 0.121 Famabus 0.075* 0.021 Poorpeer 0.247 0.016 Paratt 1.064** 0.188 Sklcmt 0.358** 0.055 Delpeer 0.475** 0.084 Delinq 0.218** -0.117 0.517** -0.032 0.135** 0.198** 0.189** 0.458 Goodness of Fit Measures: Chi-square = 12.07, df = 6, p = 0.06; CFI = 0.976; TLI = 0.889; RMSEA = 0.086. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 194

183 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 12: With delinquency (T2), internalizing, & externalizing on strain Famdis 0.443** 0.454** -0.142 0.146 Famabus 0.072 -0.053 0.106 0.036 Poorpeer 0.207 -0.095 0.550* 0.041 Delinq 0.372** 0.057 0.159 0.059 0.753** 0.055 0.293 Goodness of Fit Measures: Chi-square = 5.08, df = 2, p = 0.08; CFI = 0.957; TLI = 0.612; RMSEA = 0.106. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 195

184 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 13: With delinquency (T2) & APSD impulsivity on strain Famdis 0.418** 0.061 0.114 Famabus 0.055 0.022 0.029 Poorpeer 0.180 0.070 0.021 Delinq 0.330** 0.069 0.054 0.747** 0.058 0.297 Goodness of Fit Measures: Chi-square = 3.59, df = 2, p = 0.16; CFI = 0.976; TLI = 0.830; RMSEA = 0.076. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 196

185 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 14: With delinquency (T2), &YPI impulsivity-irresponsibility domain on strain Famdis 0.367** 0.022 0.124 Famabus 0.071 0.001 0.022 Poorpeer -0.014 0.054** 0.073 Delinq 0.263** 0.029** 0.043 0.758** 0.034 0.325 Goodness of Fit Measures: Chi-square = 3.80, df = 2, p = 0.15; CFI = 0.977; TLI = 0.839; RMSEA = 0.081. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 197

186 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 15: With delinquency (T2) & YPI impulsivity dimension on strain Famdis 0.441** 0.022 0.109 Famabus 0.079* -0.002 0.022 Poorpeer 0.114 0.082* 0.041 Delinq 0.313** 0.051** 0.057 0.772** 0.048 0.307 Goodness of Fit Measures: Chi-square = 3.60, df = 2, p = 0.16; CFI = 0.977; TLI = 0.837; RMSEA = 0.076. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 198

187 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 16: With delinquency (T2) & YPI irresponsibility dimension on strain Famdis 0.403** 0.053 0.124 Famabus 0.072 0.002 0.022 Poorpeer -0.026 0.192** 0.142 Delinq 0.324** 0.057** 0.052 0.755** 0.026 0.308 Goodness of Fit Measures: Chi-square = 3.95, df = 2, p = 0.14; CFI = 0.977; TLI = 0.840; RMSEA = 0.084. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 199

188 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 17: With delinquency (T2) & YP I thrill-seeking dimension on strain Famdis 0.386** 0.049 0.123 Famabus 0.066 0.005 0.023 Poorpeer 0.192 0.030 0.019 Delinq 0.298** 0.055** 0.042 0.760** 0.059 0.313 Goodness of Fit Measures: Chi-square = 3.62, df = 2, p = 0.16; CFI = 0.976; TLI = 0.834; RMSEA = 0.077. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 200

189 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 18: With delinquency (T2), internalizing, & externalizing on strain, social control, & delinquent peers Famdis 0.478** 0.491** -0.267 0.162 Famabus 0.073* -0.044 0.064 0.027 Poorpeer 0.221 -0.090 0.396 0.029 Paratt 1.064** 0.186 Sklcmt 0.358** 0.054 Delpeer 0.475** 0.086 Delinq 0.219** 0.010 -0.045 -0.113 0.519** -0.030 0.131** 0.204** 0.194** 0.462 Goodness of Fit Measures: Chi-square = 19.18, df = 9, p = 0.02; CFI = 0.963; TLI = 0.829; RMSEA = 0.091. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 201

190 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 19: With delinquency (T2), APSD impulsivity on strain, social control, & delinquent peers Famdis 0.458** 0.064 0.129 Famabus 0.059 0.016 0.025 Poorpeer 0.180 0.070 0.021 Paratt 1.064** 0.188 Sklcmt 0.357** 0.055 Delpeer 0.475** 0.084 Delinq 0.226** -0.012 -0.120* 0.518** -0.033 0.138** 0.199** 0.194** 0.459 Goodness of Fit Measures: Chi-square = 12.88, df = 7, p = 0.08; CFI = 0.978; TLI = 0.891; RMSEA = 0.078. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 202

191 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 20: With delinquency (T2), YPI impulsivity-irresponsibility on strain, social control, & delinquent peers Famdis 0.396** 0.025** 0.144 Famabus 0.073 0.000 0.021 Poorpeer -0.007 0.052** 0.069 Paratt 1.064** 0.186 Sklcmt 0.363** 0.057 Delpeer 0.475** 0.084 Delinq 0.181* 0.011 -0.115 0.531** -0.037 0.128** 0.188** 0.177** 0.463 Goodness of Fit Measures: Chi-square = 10.72, df = 6, p = 0.10; CFI = 0.983; TLI = 0.900; RMSEA = 0.076. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 203

192 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 21: With delinquency (T2), YPI impulsivity on strain, social control, & delinquent peers Famdis 0.482** 0.024 0.126 Famabus 0.079* -0.002 0.021 Poorpeer 0.120 0.079* 0.039 Paratt 1.064** 0.187 Sklcmt 0.365** 0.057 Delpeer 0.475** 0.084 Delinq 0.208** 0.007 -0.112 0.520** -0.032 0.133** 0.196** 0.185** 0.458 Goodness of Fit Measures: Chi-square = 10.85, df = 6, p = 0.09; CFI = 0.982; TLI = 0.893; RMSEA = 0.077. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 204

193 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 22: With delinquency (T2), YPI irresponsibility on strain, social control, & delinquent peers Famdis 0.431** 0.061* 0.144 Famabus 0.074 0.000 0.021 Poorpeer -0.022 0.190** 0.138 Paratt 1.064** 0.187 Sklcmt 0.358** 0.055 Delpeer 0.475** 0.084 Delinq 0.200** 0.021 -0.117 0.525** -0.042 0.131** 0.192** 0.183** 0.461 Goodness of Fit Measures: Chi-square = 10.95, df = 6, p = 0.09; CFI = 0.982; TLI = 0.896; RMSEA = 0.078. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment. (Continued on the next page)

PAGE 205

194 Table 17: (Continued) Exogenous Variable Endogenous Variable Endogenous Variable Delinq (T1) Internal External Impulsive Famdis Famabus Poorpeer Paratt Sklcmt Delpeer R2 Model 23: With delinquency (T2), YP I thrill-seeking on strain, soci al control, & delinquent peers Famdis 0.414** 0.056* 0.141 Famabus 0.066 0.005 0.022 Poorpeer 0.196 0.028 0.018 Paratt 1.064** 0.187 Sklcmt 0.361** 0.056 Delpeer 0.475** 0.084 Delinq 0.177** 0.030 -0.122* 0.524** -0.029 0.129** 0.194** 0.180** 0.468 Goodness of Fit Measures: Chi-square = 11.67, df = 6, p = 0.07; CFI = 0.978; TLI = 0.873; RMSEA = 0.083. Note. p< .10, ** p < .05. Delinq = delinquency; Delpeer = delinquent peers; External = externalizing behaviors; Famabus = family abuse; Famdis = family disruption; Impulsive = psychopathy impulsivity; Internal = internalizing behaviors; Paratt = parental attachment; Poorpeer = peer strain; Sklcmt = school commitment.

PAGE 206

195 An attempt was also made to complete path analyses for drug problems and drug use. Unfortunately, the data failed to satisfy goodness-of-fit st andards in the model specifying that strain at Time 2 leading to drug problems/use at Time 2 (drug problems factor: chi-square = 4.98 df = 2, p = 0.081, CFI = 0.954, TLI = 0.769, RMSEA = 0.104; drug use level: chi-square = 5.25, df = 2, p = 0.071, CFI = 0.945, TLI = 0.754, RMSEA = 0.109). The ad hoc path analyses suggested th at “state dependence” may be a viable explanation for strain and cr iminality. Youths who reporte d higher delinquency at Time 1 were more likely to also report higher le vels of both family disruption and delinquency at Time 2. Family disruption at Time 2, however, did not significantly lead to delinquency at Time 2. Implications of these findi ng are discussed in the next chapter.

PAGE 207

196 Chapter 7 Discussion General strain theory (Agnew, 1992) ha s received a respectable amount of attention over the past twenty years. In GST, Agnew has broadened the operationalization of strain, expanding the lim ited micro-social version of the Mertonian (Merton, 1968) or classic concep tualization of strain as th e disjunction between monetary and/or status aspirations and expectations to include an assortment of strain-inducing stimuli and circumstances. The scope of strain has been expanded to include stress induced by perceptions of injustice, failures to achieve contemporaneous, as well as longterm, goals, negative life events, and noxious re lationships with others. In addition, GST calls for the consideration of the magnitude, duration, frequency, chronicity, and subjectivity in the measurement of strain. Th e result is a general theory of crime that relies on social-psychological processes a nd is no longer bound by structural (Merton, 1968) and subcultural (Cloward & Ohli n, 1959, 1961; Cohen, 1955) conceptions of strain. Since its conception, Agnew and others (e.g., Agnew, 1995, 1997, 1999, 2001; Agnew et al., 2002; Broidy & Agnew, 1997; Walsh, 2000) have continued to expand upon the generalizability of this theory. GST ha s been touted as a cogent explanation for gender, community, life-course, and personal ity differences in deviance, although the empirical support for such claims remains rela tively weak or non-existent. With regard

PAGE 208

197 to individual personality diffe rences, in particular, Agnew, Brezina, Wright, and Cullen (2002) recently published a study in corporating personality tra its in a GST framework for the etiology of crime. Speci fically, certain features of personality (i.e., negative emotionality and low constraint) were determined to moderate the effect of strain on delinquency. This most recent theoretical el aboration of GST offers the opportunity to once again enhance the generali zability of the theory. The present study attempted to replicate a nd expand Agnew et al .’s (2002) recent test of GST and personality characteristics. Using a prospective, two-year longitudinal sample of 137 justice-referred adolescents (mostly first-time misdemeanor offenders), this study examined the role of internalizing behaviors, externalizing behaviors, and psychopathy as maladaptive personality ch aracteristics in conditioning the straindelinquency and strain-drug use relationships. This study expanded on that of Agnew et al. by employing longitudinal, rather than cross-sectional, methodology. In addition, supplemental analyses examined the mediati ng role of strain be tween the personalitydelinquency relationship. The initial intention of this study was to test a structural equation model of Time 1 strain and personality on Time 2 delinque ncy and drug use. Social control and differential association measures were included as competing, yet correlated, measures in the models. The decision to correlate low so cial control and differe ntial association with strain was based on Agnew’s (1992, 2001) suppos ition that such measures are associated with and strengthen the crimi nogenic effects of strain. Si x hypotheses were articulated for the SEM models for delinquency and drug use.

PAGE 209

198 First, it was hypothesized that strain would be positively related to delinquency and drug use problems. Hypothesis #1 was not supported by the present study for Time 1 strain on Time 2 delinquency or drug use. St rain at Time 1 did not significantly lead to delinquency or drug use during Time 2. The re maining hypotheses (2-6) pertained to the nature of the relationship between strain and the personality measures. It was hypothesized that strain and pers onality characteristics at Ti me 1 would have reciprocal or feedback effects (Hypothesis #2). The two-wave data, however could not properly address the issue of reciprocal effects, in part due to data limitations, but primarily due to model misspecification. Although positive sign ificant bivariate co rrelations were observed between the strain measures and personality proxy measures (see Appendix B), the SEM models could not be properly estim ated to examine the causal relationship between strain and personality (Hypothesis #4 ). Moreover, the SEM models experienced computational difficulties, despite efforts to recode the data into more manageable measures, thus preventing any empirically meaningful conclusions regarding the relationship between personality characteris tics and strain, delinquency, or drug use. This study proposed SEM models for st rain and subseque nt delinquency and strain and subsequent drug use problems. Du e to the small sample size, the SEM models needed to be parsimonious, yet empirically meaningful. As Chapter 6 reported, the SEM models suffered from estimation difficulties as a consequence of the lack of parsimony in the predicted models. Time 1 latent variables for strain a nd social control experienced temporal multicollinearity issues, which in most cases prevented estimation. While the confirmatory factor analyses of Time 1 and Time 2 measures indicated good fit statistics for the models, the SEM results suggested that the latent measures for strain and social

PAGE 210

199 control lacked appropriate c ohesion. In most cases, one item loaded disproportionately higher than the others on the latent variables. It was po ssible these reductions in the loadings for Time 2 measures could have been due to program effects on the measures. However, ANOVA and Fisher’s Exact Test s examining group differences between youths assigned to the AIW project and the control group did not reveal any significant differences in the strain, social control, differential association, personality, delinquency, or drug use problems (including a categorical measure of dr ug use level) measures. These findings suggested that the model needed to be revised and the latent variables needed to be partialled out into separate in dicators. Consequently, supplemental analyses of the personality and strain models were conducted using path analysis. Supplemental path analyses were conduc ted predicting that delinquency and drug use at Time 2 would be directly predicted by three indexes for stra in: family disruption, family abuse/neglect, and poor peer relationshi ps. The results indicated that none of the measures of strain at Time 1 predicted Ti me 2 delinquency or drug use problems. These findings supported the limited findings from the SEM analyses. According to GST (Agnew, 1992), variati on in the propensity to commit crime and other deviance (e.g., substance use) is ma inly a function of individual differences in the level of frustration or strain and the abil ity to cope with such strain. This suggests that “population heterogeneity”—differences in individual bac kground characteristics (i.e., strain) (Nagin & Fa rrington, 1992; Nagin & Pate rnoster, 1991)—provides the etiological foundation for explaini ng differences in crime. An alternative explanation for crime has been suggested by the “state dependence” approach (Nagin & Farrington, 1992; Nagin & Paternoster, 1991), which suggests th at differences that emerge as a direct

PAGE 211

200 consequence of committing the first criminal act/event (i.e., lowering the perceived costs and increasing the perceived rewards of devian ce) explain variations in crime. A few GST studies have reported findings that ma y support a state dependence explanation of GST (see Aseltine et al., 2000; Kim et al., 2003). These longitudinal studies found that delinquency was a significant predictor of strain, which was a predictor of later delinquency. A conservative test of the state depende nce explanation for GST and personality was conducted in ad hoc path analyses. The results indicated that delinquency at Time 1 was a significant predictor of strain at Time 2, and strain at Time 2 was a significant predictor of delinquency at Time 2. However, the same form of strain did not mediate the delinquency-delinquency relationship. Y ouths who had higher levels of delinquency at Time 1 were significantly more likely to report circumstances of family disruption at Time 2. Family disruption at Time 2 was not significantly related to delinquency at Time 2. Strain as measured by family abuse/negl ect at Time 2, however was significantly and positively related to delinquency at Time 2. At best, it seems prior delinquency exacerbates both subsequent strain and deli nquency. These findings support those of other longitudinal GST studies (Ase ltine et al., 2000; Kim et al., 2003). In addition to serving as a means to examine the state dependency approach for GST, the ad hoc path analyses provided an opportunity to examine the moderating effects of personality on strain. None of the models examining the effects of interactions terms between strain and personality fit the data well In this sample, the effects of strain did not appear to be moderated by levels of personality characteristics or psychopathic impulsivity indexes.

PAGE 212

201 The ad hoc path analyses were also used to examine the mediating role of strain on the personality-delinquency relationship. As discussed in Chapter 3, a substantial portion of the literature has linked maladaptiv e personality characteri stics to antisocial behavior, in particular im pulsivity (Farrington, Loeb er, & Kammen, 1990; Gerbing, Ahadi, & Patton, 1987; Krueger et al., 1994; Luengo, et al., 1994; Royce & Wiehe, 1988; White et al., 1994). Further, studies have i ndicated that personality characteristics may be somewhat inherent and stable over the life-course (for a re view see Roberts & DelVecchio, 2000). Therefore, personality traits were conceived of as predictors of strain and direct and indirect predic tors of delinquency. This c onception is consistent with Agnew’s (1992, 2001) claim that cond itioning factors (e.g., personality traits/temperament) may lead di rectly to strain as well as moderate the strain-delinquency relationship. The ad hoc results found that interna lizing and impulsivity (measured as dimensions of psychopathy in the YPI) were sign ificant predictors of strain. Youths who reported higher levels of internalizing behavior s at Time 1 were significantly more likely to experience higher levels of family di sruption at Time 2 than those with less internalizing behaviors. Youths who reported higher levels of impulsivityirresponsibility and irresponsib ility on the YPI assessment we re significantly more likely to experience higher levels of poor peer re lations than those w ith lower psychopathic features. When the direct effects of the person ality characteristics were tested on delinquency at Time 2, internalizing and ex ternalizing behaviors were not significant predictors of delinquency. All of the impul sivity psychopathy inde xes were significant

PAGE 213

202 positive predictors of subsequent delinquency. After strain measures were added to the personality-delinquency models, the relationshi ps between personality characteristics and delinquency remained much the same. When social control and differential association measures were added to the models, the effects of personality on delinquency were reduced to non-significance. Path analyses revealed that strain did not serve as a significant mediator for the person ality-delinquency relationship. Although not examined in this study, it se ems likely the indirect effects of personality on delinquency were mediated by the social control and differential association measures. Such findings may provide further validation for self-control theory (Gottfredson & Hirschi, 1990), given the impulsivity slant of the personality measures in this study. Future studies of this sample should examine the role of low social control in the effect of low self-control on delinquency. This study also examined path analyses mimicking the delinquency analyses for drug problems and drug use. The drug problem s and drug use path models did not fit the data well. Therefore, among these youths, st rain does not significantly predict drug use or drug problems. Overall, this study found little support for a general strain theory explanation of delinquency. A latent construct of stra in did not predict delinquency. Moreover, observed measures of strain as negative re lationships with peers and family disruption (e.g., fighting/disputes, criticizing) did not predict delinquency. The only strain measure that significantly led to de linquency was family abuse/negl ect. Arguably, a case can be made that this measure is also a measure of low social control. Since the data examined

PAGE 214

203 in this study did not include measures of negative affect, there was no way to know whether or not family abuse/neglect motivat ed delinquency as GST would hypothesize. This study does suggest that delinquency ex acerbates strain. Gi ven the nature of the sample, however, the relationship between self-reported delinque ncy and subsequent strain may be spurious. Since the youths in this study were all ju stice-referred, mainly first-time offenders, one can not rule out th e possibility that strain arose from the “official” attention the youth received as a result of being arrested /charged by the State Attorney’s Office and referred to attend the Juve nile Arbitration divers ion program. It is conceivable that the disrupti on caused by the youth’s official act of deviance increased strain between the youth and his/her family. Unfortunately, these data lack measures from the youths and their parents that might capture such effects. So what does this mean for the future of GST? Conceptually, the case that made in this study and by Agnew et al. (2002) that certain person ality traits should condition the effects of strain on delinquency seems vali d. Empirically, this hypothesis remained to be fully examined. Future studies are needed to examine the role of personality on strain. Limitations The models presented in this manuscrip t presented a number of limitations. The most crucial limitation was the sample size used in this study. If one follows the rule of thumb that there should be at least 10 cases pe r variable to maintain predictive power in the analyses, it becomes evident that a sample size of 137 adolescents requires the use of very parsimonious statistical models. This limitation was further compounded by desire to examine two-wave longitudinal effects, which essentially doubl es the number of variables included in the model.

PAGE 215

204 In an effort to overcome the sample size limitation, structural equation models were used. By creating latent measures, rath er than strictly observed measures, degrees of freedom are preserved and more observed measures can be included in the model. Further, exploratory and confirmatory factor analyses were used as a guide to create observed measures using multiple items. Latent measures and factor analyses provide a sound solution to measurement reduction, assuming the measures are conceptual ly cohesive. In this study, the factor analyses provided factors that contained m oderately strong loadings for most of the included items. When the SEM analyses we re conducted combining the three strain measures (family disruption, family abuse/ neglect, and poor peer relations) into one overall latent variable of strain for Time 1 and Time 2, however, the results suggested relatively weak conceptual cohesion among obs erved items. The latent measures were dominated, especially at Time 2, by the family disruption item. Another limitation of the SEM models wa s that the model was too complex to be estimated. This was also a direct function of the sample size. The moderating effects of personality on strain in the SEM models co uld not be successfully run due to the complexity of the models. When simple m odels were examined, they performed better than the more complex models. Yet, estimation problems remained. This study was also limited by the fact that it was restricted to examining measures across two periods of time. Ke ssler and Greenberg (1981) have emphasized that two waves of data may not be sufficient to provide information about how two or more variables interrelate over time. Theref ore, any findings discu ssed in this study, as well as those in the GST literature using limited longitudinal and/or cross-sectional

PAGE 216

205 analyses, should be examined critically and applied conservatively. In particular, caution should be used when interpreting the ad hoc pa th analyses of the Time 2 strain effects on Time 2 delinquency. Although Agnew (1992) argue s that strain has a rather immediate effect on delinquency, cross-sec tional applications of strain should be interpreted with prudence. Without the certainty of temporal or der, where X occurs before Y, causality is questionable. Future studies of GST should examine the effects of personality characteristics on strain and de linquency utilizing longitudinal data with three or more waves of data. Even though GST asserts that the effect s of strain on delinque ncy will be rather contemporaneous (Agnew, 1992, 2001), this does not mean that longitudinal studies are not necessary to support the empirical validity of the theory. Without a temporal order, such that strain occurs before delinquency, studies of GST can not rule out the possibility that (a) delinquency causes strain and (b) the measures may simply be correlated, but not causal. While a few studies of GST have examined longitudinal data with three or more waves of data and found support for the assumption of GST that strain leads to crime/delinquency (Aseltine et al., 20 00; Hoffman & Cerbone, 1999; Hoffmann & Miller, 1998; Kim et al., 2003), most studies have examined either cross-sectional of twowave longitudinal data. In the majority, if not almost all, of the longitudinal studies of GST the lag between waves is approximately 12 months. With such a large time span between data points, an argument can be made that the contemporaneous effects of strain on delinquency may not be detectable. Theref ore, future research on GST should attempt to collect multiple-wave longitudinal data that utilizes a smaller temporal lag between data collection points.

PAGE 217

206 Finally, this study lacked appropriate control variable (e.g., age, gender, SES, parental education, etc.) in the models. Controlling for the effects of certain sociodemographic measures, such as age and gender, may provide strikingly different results when examining delinquency and impulsiv ity. It is possible that such differences may have emerged if the path models had ex amined the influence of key covariates (e.g., gender, race/ethnicity, age) in MIMIC anal yses (see Muthn & Muthn, 2004), especially gender given that almost half of the sample was female. However, MIMIC analyses were not examined in this study. Future studies of GST and personality should control for sociodemographic characteristics that are theo retically and empirical ly associated with the key endogenous and exogenous measures. Implications Methodologically, this study emphasizes the importance of sample size when conducting research. Complex SEM models re quire the estimation of a large number of parameters in the structural and measurement models. Small samples can make estimating fairly complex models very difficult, if not impossible, as was the case in the SEM models in this study. It is possible that the estimation problems experienced with the models in this study could have been pred icted given the small sample size. Since a stepwise approach, examining a simple mode l of the main predictors then adding the personality measures if significant strain eff ects were observed, was taken in the analysis of the SEM models, however, the simple mode ls should not have experienced estimation problems based solely on sample size. However, statistical packages that are sensitive to scale differences and large variances (e.g., greater than 10) (see Muthn & Muthn, 2004) may experience

PAGE 218

207 difficulties estimating even simpler SEM models that include measures with scaling issues. In this study, the analyses began with factor scores of the st rain, social control, and differential association measures. These f actor scores were very precise, measuring constructs down to 5 decimal places. When the variables were respecified as additive summary scores, estimate was improved. Ther efore, proper specification of not only the model but also the variables is essential. In addition to methodological implications there are theoretica l implications for this study. First, the data examined in this sample suggested that in ternalizing behaviors, externalizing behaviors, and behavioral dimensions of ps ychopathy were poor moderators for the strain-delinquency rela tionships. While interaction terms of the moderating path analyses were significant, the goodness-of-fit indexes in dicated that the moderating models did not fit the data well. Agnew et al. (2002) examined the modera ting effects of a composite scale of negative emotionality and low constraint on a composite measure of strain and found significant effects for the inte raction term on delinquency. Although the interaction term was significant, the addition of this variable to the model did not reduce the coefficient sizes or effects of either stra in (without interaction: b = 0.16, B = 0.11; with interaction: b = 0.16, B = not reported ) or negative emotionality/low cons traint (without interaction: b = 0.07, B = 0.05; with interaction: b = 0.07, B = not reported ). Moreover, the interaction regression model increased th e explained variance of the model for delinquency by only 1%. Agnew and colleagues state that “the data reveal that the key personality traits of negative emotionality/low constraint condition the effect of strain on delinquency, such that strain is much more likely to lead to delinquency among those high in negative

PAGE 219

208 emotionality/low constraint.” (p. 63). Yet, they provide (a) no discussion of the significance of the slope differen ces for the simple regression of the interaction term (see Aiken & West, 1991) and (b) no standard errors for the interaction and non-interaction models. This causes one to question th e above conclusion drawn by Agnew and his colleagues. Clearly, additional research is needed on the relationship between personality characteristics and strain. The present study expanded on the Agnew et al. (2002) study by examining the effects of personality on subsequent strain, and the mediating role of strain between personality and delinquency. Maladaptive pe rsonality at Time 1 has significant effects on certain measures of strain at Time 2. Yet, strain was not found to significantly mediate the relationship between personali ty and delinquency. While personality characteristics may play an important ro le in conditioning the strain-delinquency relationship, the data examined here do not support such claims. The findings in this study stress the need for future research rega rding the influence of personality on strain and negative affect. Perhaps the lack of support for th e hypotheses in this study, and the weak support provided by Agnew et al., is due to the operationaliza tion of the strain measures. In this study and in the Agnew et al. study, strain was defined strictly as negative relationships between parents, family, peers, school, a nd others. The question arises: do these operands of strain validly reflect th e conceptualization of strain? In GST, strain is intended to reflect the motivational aspects of frustration that typically result from negative relations with others. Yet, the SEM analyses presented here suggested that for these particular data and questionnaire items, there existed much

PAGE 220

209 overlap between the strain and social control measures. The measures used in this study were based on those used by Agnew et al. (2002) in their study of GST and personality. In both cases, an argument can be made that th e strain measures are actually measures of social control. Certainly, family disr uption may also be labeled low parental commitment; family abuse may also be calle d low parental attachment; and poor peer relations or peer strain may be referred to as low peer attachment. Neither this study nor Agnew et al’s (2002) possessed measures of strain as (a) the failure to achieve positive goals and/or (b) the removal of positive stimu li. Moreover, both studies lacked measures of negative affect. As discussed in Chapter 2, a model of a critical test of GST, which may definitively permit the researcher to clai m that the measures of strain and social control represent distinct predictors of delinquency, must include intervening mechanisms. If the strain and social c ontrol measures lead to delinquency through negative affect, GST is supported and the issu e of conceptual overlap becomes mute. Future studies should make better efforts to cap ture all three types of strain and negative affect when including social control measures in models. Finally, the findings presented here ha ve significant policy implications for justice-involved youth. The ad hoc path anal yses illustrated that delinquency begets delinquency and exacerbates other negative conditions. Psychopathic behavioral characteristics significantly predict future delinquency as well, and influence certain other family and peer risk factors. Theref ore, it is essential that early intervention programs for youths involved in the juvenile justice system focus on effective strategies to improve conditions for youths and their families in a holistic fashion (Arcia, Keyes, Gallagher & Herrick, 1993; Sirles, 1 990; Tolan, Ryan & Jaffe, 1998).

PAGE 221

210 In particular, interventions that focus on family empowerment and behavioral improvements in parental interactions with their children have demonstrated substantial success in preventing recidivism and future an tisocial behavior. In a meta-analysis of family-based intervention programs, Farring ton and Welsh (2003) reported that familybased intervention programs effectively redu ced delinquency and anti social behavior by 34 to 50 percent, with long-term effects re maining for many family interventions. In particular, they found the most effective interventions were those that employed techniques to change the behavior of the parent toward the child. Along these lines, Multisystemic Treatment (MST) (Henggeler, Schoenwald, Borduin, Rowland, & Cunningham, 1998), a family-based and home-ba sed clinical approach to antisocial behavior prevention, has been shown to be effective in treating antisocial youths, including those with emotional/psychologica l functioning problems (for a recent review see Curtis, Ronan, & Borduin, 2004). In a ddition, the Family Empowerment Intervention (Dembo & Schmeidler, 2002) has shown effec tive prevention of delinquency, which is also more cost-effective than c linical interventions. As th is study intimates, family-based interventions for first-time offenders esp ecially are need to reduce recidivism and potential negative family consequences (e.g., increased family disruption and abuse/neglect) of de linquent behavior.

PAGE 222

211 References Agnew, R. (1984). Goal achievement and delinquency. Sociology and Social Research, 68 435-451. Agnew, R. (1985). A revised st rain theory of delinquency. Social Forces, 64 151167. Agnew, R. (1989). A longitudinal te st of revised strain theory. Journal of Quantitative Criminology 5 373-387. Agnew, R. (1991). Strain and subcultura l crime theory. In J. Sheley (Ed.), Criminology: A Contemporary Handbook Belmont, CA: Wadsworth. Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency. Criminology, 30 (1), 47-87. Agnew, R. (1995). The contribution of soci al-psychological stra in theory to the explanation of crime and delinquency. In F. Adler & W. S. Laufer (Eds.), Advances in Criminological Theory, Vol. 6: The Legacy of Anomie Theory New Brunswick, NJ: Transaction. Agnew, R. (1997). Stability and change in crime over the life course: A strain theory explanation. In T. P. Thornberry (Ed.), Developmental Theories of Crime and Delinquency, Advances in Criminological Theory Vol. 7. New Brunswick, NJ: Transaction. Agnew, R. (1999). A general strain theo ry of community differences in crime rates. Journal of Research in Crime and Delinquency, 36 (2), 123-155. Agnew, R. (2001). Building on the foundation of general strain theory: Specifying the types of strain most likely to lead to crime and delinquency. Journal of Research in Crime and Delinquency 38 (4), 319-361. Agnew, R. (2002). Experienced, vicarious, and anticipated strain: An exploratory study on physical victim ization and delinquency. Justice Quarterly 19 (4), 603-632. Agnew, R., & Brezina, T. (1997). Relationa l problems with peers, gender and delinquency. Youth & Society 29 84-111.

PAGE 223

212 Agnew, R., Brezina, T., Wright, J. P., & Cullen, F. T. (2002). Strain, personality traits, and delinquency: Exte nding general strain theory. Criminology 40 43-71. Agnew, R., & White, H. R. (1992). An empi rical test of general strain theory. Criminology, 30 (4), 475-499. Ahadi, S. A., & Rothbart, M. K. (1994). Temperament, development, and the Big Five. In C. F. Halverson, Jr., G. A. Kohnstamm, & R. P. Martin (Eds.), The Developing Structure of Temperament and Perso nality from Infancy to Adulthood (pp. 189-207). Hillsdale, NJ: Erlbaum. Aiken, L. S., & West, S. G. (1991). Multiple Regression: Testing and Interpreting Interactions London: Sage. Akers, R. L. (1973). Deviant Behavior: A Soc ial Learning Approach Belmont, CA: Wadsworth. Akers, R. L. (1977). Deviant Behavior: A Soc ial Learning Approach (2nd ed.). Belmont, CA: Wadsworth. Akers, R. L. (1985). Deviant Behavior: A Soc ial Learning Approach (3rd ed.). Belmont, CA: Wadsworth. Akers, R. L. (1994). Criminological Theories (2nd ed.). Los Angeles: Roxbury Publishing. Akers, R. L. (1996). Is differential associ ations/social learning cultural deviance theory? Criminology 34 229-248. Akers, R. L. (1997). Criminological Theories: Introduction and Evaluation (2nd ed.). Los Angeles, CA: Roxbury Publishing Co. Akers, R. L., & Cochran, J. K. (1985). A dolescent marijuana use: A test of three theories of deviant behavior. Deviant Behavior, 6 323-346. Alarid, L. F., Burton, V., & Cullen, F. T. (2000). Gender and crime among felony offenders. Assessing the generality of social c ontrol and differential association theories. Journal of Research in Crime and Delinquency 37 171-199. Aledort, N. (2001). Lessons from a new case management model for juvenile offenders with mental health needs. In G. Landsberg & A. Smiley (Eds.), Forensic Mental Health: Working with O ffenders with Mental Illness (pp. 331-337). Kingston, NJ: Civic Research Institute.

PAGE 224

213 Alterman, A. I., Rutherford, M. J., Cacciola, J. S., McKay, J. R., & Boardman, C. R. (1998). Prediction of 7 m onth methadone maintenance treatment responses by four measures of antisociality. Drug and Alcohol Dependence 49 217-223. Andershed, H., Kerr, M., Stattin, H., & Leva nder, S. (2002). Psychopathic traits in non-referred youths: Initial test of a new a ssessment tool. In E. Blaauw & L. Sheridan (Eds.), Psychopaths: Current Inte rnational Perspectives (pp. 131-158). The Hague: Elsevier. Anderson, E. (1999). Code of the Street New York: Norton. Arcia, E., Keyes, L., Gallagher, J. J ., & Herrick, H. (1993). National portrait of sociodemographic factors associated with underu tilization of services : Relevance to early intervention. Journal of Early Interventions, 17 283-297. Aseltine, R. H., Gore, S., & Gordon, J. (2000). Life stress, anger and anxiety, and delinquency: An empirical test of general strain theory. Journal of Health and Social Behavior 41 256-275. Asendorpf, J. B. (1992). A Brunswikean appr oach to trait con tinuity: Application to shyness. Journal of Personality 60 53-77. Asendorpf, J. B., & Aken, M. A. G. van. (1991). Correlates of the temporal consistency of personality patterns in childhood. Journal of Personality 59 689-703. Averill, J. R. (1982). Anger and Aggression: An Essay on Emotion New York: Springer-Verlag. Averill, J. R. (1993). Illusions of anger. In R. B. Felson & J. T. Tedeschi (Eds.), Aggression and Violence New York: Springer-Verlag. Babiak, P. (1995). When psychopaths go to work. International Journal of Applied Psychology 44 171-188. Babiak, P. (2000). Psychopathic manipul ation at work. In C. Gacono (Ed.), The Clinical and Forensic Assessment of Psychopathy: A Practitioner’s Guide Mahwah, NJ: Erlbaum. Bandura, A. (1973). Aggression: A Social Learning Analysis Englewood Cliffs, NJ: Prentice Hall. Bandura, A. (1977). Social Learning Theory Englewood Cliffs, NJ: Prentice Hall. Bandura, A., & Walters, R. H. (1963). Social Learning and Personality Development New York: Holt, Rinehart, and Winston.

PAGE 225

214 Bao, W.-N., Haas, A., & Pi, Y. (2004). Life strain, negative emotions, and delinquency: An empirical test of general strain theory in the People’s Republic of China. International Journal of Offender Therapy and Comparative Criminology 48 (3), 281297. Baron, S. W., & Hartnagel, T. F. (1997). Attributions, affect, and crime: Street youths’ reactions to unemployment. Criminology, 35 409-434. Baron, S. W., & Hartnagel, T. F. (2002) Street youth and labor market strain. Journal of Criminal Justice 30 519-533. Basic Behavioral Science Task Force of the National Advisory Mental Health Council (1996). Basic behavioral science res earch for mental health: Vulnerability and resilience. American Psychologist, 51 22-28. Battin, S. R., Hill, K. G., Abbott, R. D., Catalano, R. F., & Hawkins, J. D. (1998). The contribution of gang membership to delinquency beyond delinquent friends. Criminology 36 93-115. Belmore, M. F., & Quinsey, V. L. (1994). Correlates of psychopathy in a noninstitutional sample. Journal of Interpersonal Violence 9 339-349. Benda, B. B., & Corwyn, R. F. (2002). The effect of abuse in childhood and in adolescence on violence among adolescents. Youth & Society 33 (3), 339-365. Benet, V., & Walker, N. G. (1995). The big seven factor model of personality description: Evidence for its cross-cultural generality in a Spanish sample. Journal of Personality and Social Psychology 69 701-718. Benson, M. L., Fox, G. L., DeMaris, A., & Van Wyk, J. (2003). Neighborhood disadvantage, individual economic distress and violence against women in intimate relationships. Journal of Quantitative Criminology 19 (3), 207-235. Bentler, P. M. (1990). Comparative Fit Indexes in structural models. Psychological Bulletin, 107, 238-246. Berg-Nielsen, T. S., Vikan, A., & Dahl, A. A. (2002). Parenting related to child and parental psychopathology: A descri ptive review of the literature. Clinical Child Psychology and Psychiatry 7 529-552. Berkowitz, L. (1993). Aggression: Its Causes, Consequences, and Control New York: McGraw-Hill.

PAGE 226

215 Bernard, T. J. (1984). Control criticisms of strain theories: An assessment of theoretical and empirical adequacy. Journal of Research in Crime and Delinquency 21 353-372. Bernard, T. J. (1987). Testing structural strain theories. Journal of Research in Crime and Delinquency 24 262-280. Bernard, T. J. (1990). Angry aggre ssion among the truly disadvantaged. Criminology 28 73-96. Bernburg, J. G., & Thorlindsson, T. (1999). Adolescent violence, social control, and the subculture of delinquency: Factors related to violence behavior and nonviolent delinquency. Youth & Society 30 445-460. Binder, A. (1988). Juvenile delinquency. Annual Review of Psychology 39 252282. Blackburn, R., & Coid, J. W. (1998). Psychopathy and the dimensions of personality disorder in violent offenders. Personality and Individual Differences 25 129145. Blair, R. J. R. (1999). Responsiveness to distress cues in the child with psychopathic tendencies. Personality and Individual Differences 27 135-145. Blair, R. J. R., Monson, J., & Freder ickson, N. (2001). Moral reasoning and conduct problems in children with emo tional and behavioural difficulties. Personality and Individual Differences 31 799-811. Block, J. (1993). Studying personality the long way. In D. Funder, R. D. Parke, C. Tomlinson-Keasy, & K. Widaman (Eds.), Studying Lives through Time: Personality and Development (pp. 9-41). Washington, DC: Am erican Psychological Association. Block, J. (1995). A contrarian view of the five-factor approach to personality description. Psychological Bulletin 117 187-215. Block, J., Block, J. H., & Keyes, S. (1988) Longitudinal foretelling drug usage in adolescence: Early childhood personal ity and environmental precursors. Child Development 59 336-355. Block, J., & Kremen, A. (1996). IQ a nd ego-resiliency: Clarifying their conceptual and empirical linkage and separateness. Journal of Personality and Social Psychology 70 349-361.

PAGE 227

216 Blonigen, D. M., Carlson, S. R., Krueger, R. F., & Patrick, C. J. (2003). A twin study of self-reported psychopa thic personality traits. Personality and Individual Differences 35 179-197. Boardman, J. D., Finch, B. K., Ellison, C. G., Williams, D. R., & Jackson, J. S. (2001). Neighborhood disadvantage, stre ss, and drug use among adults. Journal of Health and Social Behavior 42 151-165. Bock, G. R., & Goode, J. A. (Eds.). (1996). Genetics of Criminal and Antisocial Behaviour Toronto: Wiley. Bollen, K. A. (1989). Structural Equations wi th Latent Variables New York: John Wiley & Sons. Bolt, D. M., Hare, R. D., Vitale, J. E., & Newman, J. P. (2004). A multigroup item response theory analysis of the Psychopathy Checklist—Revised. Psychological Assessment 16 (2), 155-168. Borkenau, P. (1990). Traits as ideal-base d and goal-derived social categories. Journal of Personality and Social Psychology 58 381-396. Brandt, J. R., Kennedy, W. A., Patrick, C. J., & Curtin, J. J. (1997). Assessment of psychopathy in a population of incarcerated adolescent offenders. Psychological Assessment 9 429-435. Brebner, J. (1998). Happiness and personality. Personality and Individual Differences 25 279-296. Brebner, J. (2001). Personality and stress coping. Personality and Individual Differences 31 317-327. Brezina, T. (1996). Adapting to strain: An examination of delinquent coping responses. Criminology 24 (1), 39-60. Brezina, T. (1998). Adolescent maltreatme nt and delinquency: The question of interviewing process. Journal of Research in Crime and Delinquency 26 (1), 71-99. Brezina, T. (1999). Teenage violence toward parents as an adaptation to family strain. Youth and Society 30 416-444. Broidy, L. M. (2001). A test of general strain theory. Criminology, 39 9-35. Broidy, L., & Agnew, R. (1997). Gender a nd crime: A general strain theory perspective. Journal of Research in Crime and Delinquency, 34 275-306.

PAGE 228

217 Brook, J. S., Nomura, C., & Cohen, P. (1989). A network of influences on adolescent drug involvement: Ne ighborhood, school, peer, and family. Genetic, Social, and General Psychology Monographs 115 123-145. Brook, J. S., Cohen, P., Whiteman, M., & Gordon, A. S. (1992). Psychosocial risk factors in the transition from m oderate to heavy use or abuse of drugs. In M. D. Glantz & R. W. Pickens (Eds.), Vulnerability to Drug Abuse (pp. 359-388). Washington, DC: American Psychological Association. Browne, M. W., & Cudeck, R. (1993). Altern ative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing Structural Equation Models (pp. 445-455). Long. Newbury Park, CA: Sage. Burgess, R. L., & Akers, R. L. (1966). A differential association-reinforcement theory of criminal behavior. Social Problems 14 128-147. Burton, V. S. (1991). Explaining Adult Criminality: Testing Strain, Differential Association, and Control Theories Unpublished dissertation, Un iversity of Cincinnati, Ann Arbor, MI: University Microfilms. Burton, V. S., & Cullen, F. T. (1992). The empirical status of strain theory. Crime and Justice 15 1-30. Burton, V. S., Cullen, F. T., Evans, T. D., & Dunaway, R. G. (1994). Reconstructing strain theory: Operationalization, rival theo ries, and adult criminality. Journal of Quantitative Criminology 10 213-239. Buss, A. H., & Plomin, R. (1975). Temperament: Early Developing Personality Traits Hillsdale, NJ: Erlbaum. Byrne, B. M. (2001). Structural Equations Modeling with AMOS: Basic Concepts, Applications, and Programming Mahwah, NJ: Lawrence Erlbaum Associates. Campbell, M. A., Porter, S., & Santor, D. (2004). Psychopathic traits in adolescent offenders: An evaluation of cr iminal history, clinical, and psychosocial correlates. Behavioral Sciences and the Law 22 23-47. Capowich, G. E., Mazerolle, P., & Piquer o, A. (2001). General strain theory, situational anger, and social networks: An assessment of conditioning influences. Journal of Criminal Justice 29 445-461. Carey, G., & Goldman, D. (1997). The genetics of antisocial behavior. In D. M. Stoff & J. Breiling (Eds.), Handbook of Antisocial Behavior New York: Wiley.

PAGE 229

218 Carmichael, C. M., & McGue, M. (1 994). A longitudinal family study of personality change and stability. Journal of Personality 62 1-20. Carvajal, S. C., Clair, S. D., Nash, S. G., & Evans, R. I. (1998). Relating optimism, hope, and self-esteem to social influences in deterring substance use in adolescents. Journal of Social and Clinical Psychology 17 443-465. Caspi, A., Begg, D., Dickson, N., Harrington, H. L., Lanlgey, J., Moffitt, T. E., & Silva, P. A. (1997). Personality differen ces predict health-risk behaviors in young adulthood: Evidence from a longitudinal study. Journal of Personality and Social Psychology 73 (5), 1052-1063. Caspi, A., & Bem, D. J. (1990). Personal ity continuity and ch ange across the life course. In L. A. Pervin (Ed.), Handbook of Personality: Theory and Research New York: Guilford. Caspi, A., Elder, G. H., & Bem, D. J. (1988). Moving away from the world: Lifecourse patterns of shy children. Developmental Psychology 24 824-831. Caspi, A., Moffitt, T. E., Silva, P. A., Stouthamer-Loeber, M., Krueger, R. F., & Schmutte, P. S. (1994). Are some people crim e-prone? Replications of the personalitycrime relationship across countries, genders, races, and methods. Criminology 32 (2), 163-195. Caspi, A., & Roberts, B. W. (1999). Pers onality continuity a nd change across the life course. In L. A. Pervin & O. P. John (Eds.), Handbook of Personality Psychology: Theory and Research (2nd ed., pp. 300-326). New York: Guilford Press. Caspi, A., & Silva, P. A. (1995). Te mperamental qualities at age 3 predict personality traits in young adulthood: L ongitudinal evidence from a birth cohort. Child Development 66 486-498. Catchpole, R. E. H., & Gretton, H. M. (2003). The predictive validity of risk assessment with violent young offenders: A 1year examination of criminal outcome. Criminal Justice and Behavior 30 (6), 688-708. Cernkovich, S. N., & Giodano, P. C. (1979). Delinquency, opportunity, and gender. Journal of Criminal Law and Criminology, 70 145-151. Chamlin, M. B., & Cochran, J. K. (1995) Assessing Messner and Rosenfeld’s institutional anomie theory: A partial test. Criminology, 33 411-429. Chassin, L., Pillow, D. R., Curran, P. J., Molina, B. S. G., & Barrera, M., Jr. (1993). Relation of parental alcoholism to early adolescent substance us e: A test of three mediating mechanisms. Child Development 66 486-498.

PAGE 230

219 Church, A. T. (1994). Relating the Tellegen and five-factor models of personality structure, Journal of Personality and Social Psychology 67 898-909. Clark, D. B., Parker, A. M., & Lync h, K. G. (1999). Psychopathology and substance-related problems during early adolescence: A survival analysis. Journal of Clinical Child Psychology 28 333-341. Clausen, J. A. (1993). American Lives: Looking Back at the Children of the Great Depression New York: Free Press. Cleckley, H. (1941). The Mask of Sanity St. Louis: Mosby. Cleckley, H. (1976). The Mask of Sanity: An Attemp t to Clarify Some Issues about the So-Called Psychopathic Personality (5th ed.). St. Louis: Mosby. Cleckley, H. (1982). The Mask of Sanity (Rev. ed.). St. Louis: Mosby. Cloninger, C. R. (1987). A systematic method for clinical description and classification of personality variants. Archives of General Psychiatry 44 573-588. Cloninger, C. R., Svrakic, D. M., & Pr zybeck, T. R. (1993). A psychobiological model of temperament and character. Archives of General Psychiatry 50 975-990. Cloward, R. (1959). Illegitimate means, anomie, and deviant behavior. American Sociological Review, 24 (2), 164-179. Cloward, R., & Ohlin, L. (1959). Illegitimate means, anomie, and delinquent behavior. American Sociological Review, 24 164-177. Cloward, R., & Ohlin, L. (1961). Delinquency and Opportunity Glencoe, IL: Free Press. Cohen, A. K. (1955). Delinquent Boys Glencoe, IL: Free Press. Cohen, A. K. (1965). The sociology of the deviant act: Anomie theory and beyond. American Sociologi cal Review, 30 (1), 5-14. Cohen, P. (1996). Childhood risks for young adult symptoms of personality disorder: Method and substance. Multivariate Behavioral Research 31 121-148. Colder, C. R., & Chassin, L. (1997). Af fectivity and impulsivity: Temperament risk for adolescent alcohol involvement. Psychology of Addictive Behaviors 11 83-97.

PAGE 231

220 Cole, S. (1975). The growth of scientific knowledge: Theories of deviance as a case study. In L. Coser (Ed.), The Idea of Social Structure: Papers in Honor of Robert K. Merton New York: Harcourt Brace Jovanovich. Colvin, M. (2000). Crime and Coercion New York: St. Martin’s Press. Compton, W. M., Helzer, J. E., Hwu, H. G., Yeh, E. K., McEvoy, L., Tipp, J. E., & Spitznagel, E. L. (1991). New methods in cr oss-cultural psychiatry : Psychiatric illness in Taiwan and the United States. American Journal of Psychiatry 148 1697-1704. Conrod, P. J., Pihl, R. O., & Vassileva, J. (1998). Differential sensitivity to alcohol reinforcement in groups of men at risk for distinct alcoholism subtypes. Alcoholism: Clinical and Experimental Research 22 585-597. Cook, F. (1992). TASC: Case Management Models Linking Criminal Justice and Treatment. In R. S. Ashery (Ed.), Progress and Issues in Case Management Rockville, MD: National Institute on Drug Abuse. Cook, F. (2002). Treatment Accountability for Safer Communities: Linking the criminal justice and treatment systems. In C. G. Leukefeld, F. Tims, & D. Farabee (Eds.), Treatment of Drug Offenders: Policies and Issues (pp. 105-110). New York: Springer Publishing Co. Cooke, D. J. (1997). Psychopaths: Over sexed, overplayed, but not over here? Criminal Behaviour and Mental Health 7 3-11. Cooke, D. J., Kosson, D., & Michie, C. (2001). Psychopathy and ethnicity: Structural, item, and test generalizability of the Psychopathy Checklist—Revised (PCLR) in Caucasian and African American participants. Psychological Assessment 13 531542. Cooke, D. J., & Michie, C. (1999). An item response theory analysis of the Hare Psychopathy Checklist-Revised. Psychological Assessment 9 3-14. Cooke, D. J., & Michie, C. (2001). Re fining the construct of psychopathy: Towards a hierarchical model. Psychological Assessment 13 171-188. Cooper, M. L., Frone, M. R., Russell, M., & Mudar, P. (1995). Drinking to regulate positive and negative emotions: A motivational model of alcohol use. Journal of Personality and Social Psychology 69 990-1005. Corrado, R. R., Vincent, G. M., Hart, S. D., & Cohen, I. M. (2004). Predictive validity of the Psychopathy Check list: Youth Version for gene ral and violent recidivism. Behavioral Sciences and the Law 22 5-22.

PAGE 232

221 Costa, P. T. Jr., Somerfield, M. R., & McCrae, R. R. (1996). Personality and coping: A reconceptualization. In M. Zeidner & N. S. Endler (Eds.), Handbook of Coping: Theory, Research, Applications New York: Wiley. Cullen, F. T. (1984). Rethinking Crime and Deviance Theory: The Emergence of a Structural Tradition Totowa, NJ: Rowman and Allanheld. Curtis, N. M., Ronan, K. R., & Borduin, C. M. (2004). Multisystemic Treatment: A meta-analysis of outcome studies. Journal of Family Psychology, 18 (3), 411-419. Dembo, R., & Brown, R. (1994). The Hillsborough County Juvenile Assessment Center. Journal of Child and Adolescent Substance Abuse, 3 25-43. Dembo, R., Getreu, A., Williams, L., Berry, E., La Voie, L., Genung, L., Schmeidler, J., Wisk, E. D., & Kern, J. ( 1990). A longitudinal study of the relationships among alcohol use, marijuana/hashish use, cocaine use, and emotional/psychological functioning problems in a cohort of high-risk youths. The International Journal of the Addictions 25 1341-1382. Dembo, R., & Schmeidler, J. (2002). Family Empowerment Intervention: Innovative Service for High-Risk Youths and Their Families New York: Haworth Press. Dembo, R., Wareham, J., Poythress, N ., Cook, B., & Schmeidler, J. (2004a). The Impact of Arbitration Intervention Serv ices on Arbitration Program Completion Unpublished manuscript, University of South Florida at Tampa. Dembo, R., Wareham, J., Poythress, N ., Cook, B., & Schmeidler, J. (2004b). The Impact of Arbitration Intervention Services on Psychosocial Func tioning: A Follow-Up Study Unpublished manuscript, University of South Florida at Tampa. Dembo, R., Wareham, J., Poythress, N ., Cook, B., & Schmeidler, J. (2004c). The Impact of Arbitration Interventi on Services on Youth Recidivism. Unpublished manuscript, University of South Florida at Tampa. Dembo, R., Wareham, J., Poythress, N., Cook, B., & Schmeidler, J. (in press). The correlates and consequences of drug i nvolvement among youths entering a juvenile justice diversion program. Journal of Child & Adolescent Substance Abuse Dembo, R., Wareham, J., Poythress, N., Meyers, K., Cook, B., & Schmeidler, J. (under review). Continuities in prob lem behavior among high risk youths. Journal of Child & Adolescent Substance Abuse

PAGE 233

222 Dembo, R., Williams, L., Schmeidler, J., Be rry, E., Werner, W., Getreu, A., Wish, E. D., & Christensen, C. (1992a). A structur al model examining the relationship between physical child abuse, sexual victimizati on, and marijuana/hashish use in delinquent youth: A longitudinal study. Violence and Victims 1 41-62. Dembo, R., Williams, L., Wernre, W., Sc hmeidler, J., & Brown, C. H. (1992b). The role of family factors, physical abuse, and sexual victimization experiences in highrisk youths’ alcohol and ot her drug use and delinquenc y: A longitudinal model. Violence and Victims 3 245-266. Digman, J. M., & Shmelyov, A. G. (1996). The structure of temperament and personality in Russian children. Journal of Personality and Social Psychology 71 341351. Dishion, T. J., Patterson, G. R., Stoolmille r, M., & Skinner, M. L. (1991). Family, school, and behavioral antecedents to early a dolescent involvement w ith antisocial peers. Developmental Psychology 27 (1), 172-180. Dobkin, P. L., Tremblay, R. E., Masse, L. C., & Vitaro, F. (1995). Individual and peer characteristics in pred icting boys’ early onset of s ubstance abuse: A seven-year longitudinal study. Child Development 66 1198-1214. Dohrenwend, B. P. (1998). Adversity, Stress, and Psychopathology New York: Oxford University Press. Dohrenwend, B. P. (2000). The role of adversity and stress in psychopathology: Some evidence and its implications for theory and research. Journal of Health and Social Behavior 41 1-19. Durkheim, E. (1951). Suicide: A Study in Sociology Glencoe, IL: Free Press. (Original work published 1897) Durkheim, E. (1964). The Division of Labor New York: Free Press. (Original work published 1893) Edens, J. F., Skeem, J. L., Cruise, K. R ., & Cauffman, E. (2001). Assessment of “juvenile psychopathy” and its associati on with violence: A critical review. Behavioral Sciences and the Law 19 53-80. Eitle, D. J. (2002). Exploring a source of deviance-producing strain for females perceived discrimination and general strain theory. Journal of Criminal Justice 30 429442.

PAGE 234

223 Eitle, D., & Turner, R. J. (2002). Exposure to community violence and young adult crime: The effects of witnessing vi olence, traumatic victimization, and other stressful life events. Journal of Research in Crime and Delinquency 39 (2), 214-237. Eitle, D., & Turner, R. J. (2003). Stress exposure, race, and young adult male crime. The Sociological Quarterly 44 (2), 243-269. Eley, T. C. (1998). General genes: A new theme in developmental psychopathology. Current Directions in Psychological Science 6 90-95. Elliott, A. J., & Church, M. A. (1997). A hierarchical model of approach and avoidance achievement motivation. Journal of Personality and Social Psychology 72 218-232. Elliott, D. S., Ageton, S. S., Huizinga, D., Knowles, B. A., & Canter, R. J. (1983). The Prevalence and Incidence of Delinquent Behavior: 1976-1980 Boulder, CO: Behavioral Research Institute. Elliott, D. S., & Huizinga, D. (1983). Social class and delinquent behavior in a national youth panel. Criminology, 21 149-177. Elliott, D. S., Huizinga, D., & Ageton, S. S. (1985). Explaining Delinquency and Drug Use Beverly Hills, CA: Sage. Elliott, D. S., Huizinga, D., & Menard, S. (1989). Multiple Problem Youth: Delinquency, Substance Use, and Mental Health Problems New York: Springer-Verlag. Eysenck, H. J. (1977). Crime and Personality London: Routledge & Kegan Paul. Eysenck, H. J. (1992). Four ways five factors are not basic. Personality and Individual Differences 13 667-673. Eysenck, H. J., & Eysenck, M. (1985). Personality and Individ ual Differences: A Natural Science Approach New York: Plenum. Eysenck, H. J., & Eysenck, S. B. G. (1976). Psychoticism as a Dimension of Personality London: Hodder & Stoughton. Eysenck, H. J., & Gudjonsson, G. (1989). The Causes and Cures of Criminality New York: Plenum Press. Eysenck, M. W. (1989). Personality, stress arousal, and cognitive processes in stress transactions. In R. W. Neufeld (Ed.), Advances in Investi gation of Psychological Stress (pp. 133-160). Chichest er, England: Wiley.

PAGE 235

224 Falkenbach, D. M., Poythress, N. G ., & Heide, K. M. (2003). Psychopathic features in a juvenile diversi on population: Reliability and pred ictive validity of two selfreport measures. Behavioral Sciences and the Law 21 787-806. Farnworth, M., & Leiber, M. J. (1989). St rain theory revisited: Economic goals, educational means, and delinquency. American Sociological Review 54 263-274. Farrington, D. P. (1986). Stepping stones to adult criminal careers. In D. Olweus, J. Block, & M. Radke-Yarrow (Eds.), Development of Antisocial and Prosocial Behavior (pp. 359-384). Orlando, FL: Academic Press. Farrington, D. P. (1992). Juvenile delinquency. In J. Coleman (Ed.), The School Years: Current Issues in th e Socialization of Young People (2nd ed., pp. 123-163). London: Routledge. Farrington, D. P., Loeber, R., & Kammen, W. B. van. (1990). Long-term criminal outcomes of hyperactivity-impulsivity-att ention deficit and conduct problems in childhood. In L. N. Robins & M. Rutter (Eds.), Straight and Devious Path Ways from Childhood to Adulthood (pp. 62-81). Cambridge: Un iversity Press. Farrington, D. P., & Welsh, B. C. (2003). Family-based prevention of offending: A meta-analysis. Australian and New Zealand Jo urnal of Criminology, 36 (2), 127-151. Fergusson, D., & Lynskey, M. T. (1998) Conduct problems in childhood and psychosocial outcomes in young adulthood: A prospective study. Journal of Emotional and Behavioral Disorders 6 2-18. Fergusson, D., Lynskey, M. T., & Horw ood, L. J. (1994). Alcohol consumption and associated problems in a birth cohort of 15 year olds. New Zealand Journal of Medicine 107 167-170. Fergusson, D. M., & Horwood, L. J. ( 1996). The role of adolescent peer affiliations in the continuity between ch ildhood behavioral adjustment and juvenile offending. Journal of Abnormal Child Psychology, 24 205-221. Fergusson, D. M., Woodward, L. J., & Horwood, L. J. (1999). Childhood peer relationship problems and young pe ople’s involvement with devi ant peers in adolescence. Journal of Abnormal Child Psychology, 27 357-369. Forth, A. E. (1995). Psychopathy and Young Offenders: Prevalence, Family Background, and Violence Program Branch Users Report. Ottawa: Ministry of the Solicitor General of Canada.

PAGE 236

225 Forth, A. E., & Burke, H. C. (1998). Psychopathy in adolescence: Assessment, violence, and developmental precursors. In D. Cooke, A. E. Forth, & R. D. Hare (Eds.), Psychopathy: Theory, Research, and Implications for Society Dordrecht: Kluwer. Forth, A. E., Hart, S. D., & Hare, R. D. (1990). Assessment of psychopathy in male young offenders. Psychological Assessment: A Jour nal of Consulting and Clinical Psychology 2 1-3. Forth, A. E., Kosson, D. S., & Hare, R. (2003). Hare Psychopathy Checklist: Youth Version Toronto, Ontario: Multi-Health Systems. Forth, A. E., & Tobin, F. (1995). Psychopathy and young offenders: Rates of childhood maltreatment. Forum for Corrections Research 7 20-27. Frick, P. J. (2000). The problems of inte rnal validation without a theoretical context: The different conceptual underp innings of psychopathy and the disruptive behavior disorder criteria. Psychological Assessment 12 (4), 451-456. Frick, P. J. (2002). Juvenile psychopat hy from a developmental perspective: Implications for construct developmen t and use in forensic assessments. Law and Human Behavior 26 (2), 247-253. Frick, P. J., Bodin, S. D., & Barry, C. T. (2000). Psychopathic traits and conduct problems in community and clinic-referred sa mples of children: Further development of the Psychopathy Screening Device. Psychological Assessment 12 382-393. Frick, P. J., Cornell, A. H., Barry, C. T., Bodin, S. D., & Dane, H. E. (2003). Callous-unemotional traits and conduct problems in the pr ediction of conduct severity, aggression, and self-report of delinquency. Journal of Abnormal Child Psychology 31 (4), 457-470. Frick, P. J., & Hare, R. D. (2001). The Antisocial Process Screening Device Toronto: Multi-Health Systems, Inc. Frick, P. J., Lilienfeld, S. O., Ellis, M., Loney, B., & Silverthorn, P. (1999). The association between anxiety and ps ychopathy dimensions in children. Journal of Abnormal Child Psychology 27 383-392. Frick, P. J., O’Brien, B. S., Wootton, J. M., & McBurnett, K. (1994). Psychopathy and conduct problems in children. Journal of Abnormal Psychology 103 700-707. Furnham, A. (1984). Personali ty, social skills, anomie and delinquency: A selfreport study of a group of norma l non-delinquent adolescents. Journal of Child Psychology and Psychiatry 25 409-420.

PAGE 237

226 Furnham, A., & Heaven, P. (1999). Personality and Social Behaviour London: Arnold. Furnham, A., & Thompson, J. (1991). Pers onality and self-rep orted delinquency. Personality and Individual Differences 12 585-593. Ge, X., & Conger, R. D. (1999). Adjust ment problems and emerging personality characteristics from early to late adolescence. American Journal of Community Psychology 27 (3), 429-459. Gerbing, D. W., Ahadi, S. A., & Patton, J. H. (1987). Toward a conceptualization of impulsivity: Components across the behavioral and self-report domains. Multivariate Behavioral Research 22 357-379. Giancola, P. R., & Parker, A. M. (2001). A six year prospective study of pathways toward drug use in adolescent boys with and wi thout a family history of a substance use disorder. Journal of Studies on Alcohol 62 (2), 166-178. Gillmore, M. R., Hawkins, J. D., Catalano, R. F., Day, L. E., Moore, M., & Abbott, R. (1991). Structure of pr oblem behaviors in preadolescence. Journal of Consulting Clinical Psychology 59 499-506. Godley, S. H., Finch, M., Dougan, L., McDonnell, M., McDermeit, M., & Carey, A. (2000). Case management for dually diagnos ed individuals involved in the criminal justice system. Journal of Substance Abuse Treatment, 18 (2), 137-148. Goldsmith, H. H. (1996). Studying temperam ent via construction of the Toddler Behavior Assessment Questionnaire. Child Development 67 218-235. Gonalves, R. A. (1999). Psyc hopathy and offender types. International Journal of Law and Psychiatry 22 337-346. Gottesman, I., & Goldsmith, H. H. ( 1994). Developmental psychopathology of antisocial behavior: Inserting genes into its ontogenesis and epig enesis. In C. Nelson (Ed.), Threats to Optimal Development: Integrating Biological, Psychological, and Social Risk Factors Hillsdale, NJ: Lawrence Erlbaum. Gottfredson, M. R., & Hirschi, T. (1990). A General Theory of Crime Stanford, CA: Stanford University Press. Gough, H. G., & Peterson, D. R. (1952). Th e identification a nd measurement of predispositional factors in crime and delinquency. Journal of Consulting Psychology 16 207-212.

PAGE 238

227 Gretton, H. M., Hare, R. D., & Catch pole, R. E. H. (2004). Psychopathy and offending from adolescence to adulthood: A 10-year follow-up. Journal of Consulting and Clinical Psychology, 72 (4), 636-645. Gretton, H. M., McBride, M., Hare, R. D., O’Shaughnessy, R., & Kumka, G. (2001). Psychopathy and recidivism in adolescent sex offenders. Criminal Justice and Behavior 28 427-449. Haan, N., Millsap, R., & Hartka, E. (1986). As time goes by: Change and stability in personality over 50 years. Psychology and Aging 1 220-232. Hagan, J., & Foster, H. (2003). S/He’s a re bel: Toward a sequential stress theory of delinquency and gendered pathways to disadvantage in early adulthood. Social Forces 82 (1), 53-86. Hare, R. D. (1980). A research scale for the assessment of psychopathy in criminal populations. Personality and Individual Differences 1 111-119. Hare, R. D. (1981). Psychopathy and violence In J. R. Hayes, T. K. Roberts, & K. S. Solway (Eds.), Violence and the Vi olent Individual Jamaica, NY: Spectrum. Hare, R. D. (1991). The Hare Psychopathy Checklist-Revised Toronto: MultiHealth Systems. Hare, R. D. (1993). Without Conscience: The Disturbing World of the Psychopaths among Us New York: The Guilford Press. Hare, R. D. (1996). Psychopathy: A clin ical construct whose time has come. Criminal Justice and Behavior 23 25-54. Hare, R. D. (1998a). Psychopaths and thei r nature: Implications for the mental health and criminal justice systems. In T. Mi llon, E. Simonsen, M. Birket-Smith, & R. D. Davis (Eds.), Psychopathy: Antisocial, Crim inal, and Violent Behavior (pp. 188-212). New York: Guilford. Hare, R. D. (1998b). The Hare PCL-R: Some issues concerning its use and misuse. Legal and Criminological Psychology 3 99-119. Hare, R. D., Clark, D., Grann, M., & Thornton, D. (2000). Psychopathy and the predictive validity of the PCL-R: An international perspective. Behavioral Sciences and the Law 18 623-645. Hare, R. D., & Jutai, J. W. (1983). Crimin al history of the male psychopath: Some preliminary data. In K. T. Van Dusen & S. A. Mednick (Eds.), Prospective Studies of Crime and Delinquency Boston: Kluwer-Nijhoff.

PAGE 239

228 Hare, R. D., & McPherson, L. M. (1984) Violent and aggres sive behavior by criminal psychopaths. International Journal of Law and Psychiatry 7 35-50. Hare, R. D., McPherson, L. M., & Fort h, A. E. (1988). Male psychopaths and their criminal careers. Journal of Consulting and Clinical Psychology 56 710-714. Hare, R. D., & Neumann, C. S. (2005) Structural models of psychopathy. Current Psychiatry Reports 7 Harpur, T. J., Hakstian, A. R., & Hare, R. D. (1988). Factor structure of the Psychopathy Checklist. Journal of Consulting and Clinical Psychology 56 741-747. Harpur, T. J., Hare, R. D., & Ha kstian, A. R. (1989). Two-factor conceptualization of psychopathy: Constr uct validity and assessment implications. Psychological Assessment 1 6-17. Harpur, T. J., Hart, S. D., & Hare, R. D. (1994). Personality of the psychopath. In P. T. Costa, Jr. and T. Widiger (Eds.), Personality Disorders and the Five Factor Model of Personality Washington, D.C.: American Psychological Association. Harris, G. T., Rice, M. E., & Cormier, C. A. (1991). Psychopathy and violent recidivism. Law and Human Behavior 15 223-236. Harris, G. T., Rice, M. E., & Quinsey, V. L. (1994). Psychopathy as a taxon: Evidence that psychopaths are a discrete class. Journal of Consulting and Clinical Psychology 62 387-397. Harris, G. T., Skilling, T. A., & Rice, M. E. (2001). The construct of psychopathy. In M. Tonry (Ed.), Crime and Justice: A Review of Research (pp. 197-264). Chicago: The University of Chicago Press. Hart, S. D., & Hare, R. D. (1989). Di scriminant validity of the Psychopathy Checklist in a forensic psychiatric population. Psychological Assessment: A Journal of Consulting and Clinical Psychology 1 211-218. Hart, S. D., & Hare, R. D. (1994). Ps ychopathy and the Big 5: Correlations between observers’ ratin gs of normal and pathological personality. Journal of Personality Disorders 8 32-40. Hart, S. D., Kropp, P. R., & Hare, R. D. (1988). Performance of male psychopaths following conditional release from prison. Journal of Consulting and Clinical Psychology 56 227-232. Hart, S. D., Watt, K. A., & Vincent, G. M. (2002). Commentary on the Seagrave and Grisso: Impressions of the state of the art. Law and Human Behavior 26 (2), 241-245.

PAGE 240

229 Hay, C. (2003). Family strain, gender, and delinquency. Sociological Perspectives 46 (1), 107-135. Haynie, D. L. (2001). Delinquent peers revi sited: Does network structure matter? American Journal of Sociology, 106 1013-1057. Heaven, P. C. L. (1996). Personality and self-reported delinquency: An analysis of the “big five” personality dimensions. Personality and Individual Differences 20 4754. Heilbrun, K., Hart, S. D., Hare, R. D., Gu stafson, D., Nunez, C., & White, A. J. (1998). Inpatient and postdischar ge aggression in mentally di sordered offenders: The role of psychopathy. Journal of International Violence 13 515-527. Helson, R., Stewart, A. J., & Ostrove, J. (1995). Identity in three cohorts of midlife women. Journal of Personality and Social Psychology 69 544-557. Hemphill, J. F., Hare, R. D., & Wong, S. (1998). Psychopathy and recidivism: A review. Legal and Criminological Psychology 3 139-170. Hemphill, J. F., Hart, S. D., & Hare, R. D. (1994). Psychopathy and substance use. Journal of Personality Disorders 8 169-180. Henggeler, S. W., Schoenwald, S. K ., Borduin, C. M., Rowland, M. D., & Cunningham, P. B. (1998). Multisystemic Treatment of Antisocial Behavior in Children and Adolescents New York: Guilford Press. Herrenkohl, T. I., Maguin, E., Hill, K. G ., Hawkins, J. D., Abbott, R. D., & Catalano, R. F. (2000). Developmenta l risk factors for youth violence. Journal of Adolescent Health 26 176-186. Hicks, M. M., Rogers, R., & Cashel, M. (2000). Predictions of violent and total infractions among institutionalized male juvenile offenders. Journal of the American Academy of Psychiatry and the Law 28 183-190. Hill, C. D., Rogers, R., & Bickford, M. E. (1996). Predicting aggressive and socially disruptive behavior in a maximum security forensic psychiatric hospital. Journal of Forensic Sciences 41 56-59. Hindelang, M., Hirschi, T., & Weiss, J. (1981). Measuring Delinquency Beverly Hills: Sage. Hirschi, T. (1969). Causes of Delinquency Berkeley, CA: University of California Press.

PAGE 241

230 Hobson, J., & Shine, J. (1998). Measurem ent of psychopathy in a U.K. prison population referred for long-term psychotherapy. British Journal of Criminology 38 504-515. Hobson, J., Shine, J., & Roberts, R. (2000). How do psychopaths behave in a therapeutic community? Psychology, Crime and Law 6 139-154. Hoffmann, J. P. (2002). A contextual analys is of differential association, social control, and strain th eories of delinquency. Social Forces 81 (3), 753-785. Hoffmann, J. P., & Cerbone, F. G. (1999). Stressful life events and delinquency escalation in early adolescence. Criminology 37 (2), 343-373. Hoffmann, J. P., & Ireland, T. O. (2004) Strain and opportunity structures. Journal of Quantitative Criminology 20 (3), 263-292. Hoffmann, J. P., & Miller, A. S. (1998). A latent variable analysis of general strain theory. Journal of Quantitative Criminology 14 (1), 83-110. Hoffmann, J. P., & Su, S. S. (1997). The conditional effects of stress on delinquency and drug use: A strain theo ry assessment of sex differences. Journal of Research in Crime and Delinquency 34 (3), 275-306. Hofstee, W. K. B. (1994). Will the tr ue trait theorist please stand up? Psychological Inquiry 5 134-137. Hogan, R. (1986). Hogan Personality Inventory Minneapolis, MO: National Computer System. Hollingshead, A., & Redlich, F. C. (1958). Social Class and Mental Illness: A Community Study New York: Wiley. Hu, L.-T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conv entional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1-55. Hughes, G., Hogue, T. E., Hollin, C. R., & Champion, H. (1997). First stage evaluation of a treatment programme for personality disordered offenders. Journal of Forensic Psychiatry 8 515-527. Hume, M. P., Kennedy, W. A., Patric k, C. J., & Partyka, D. J. (1996). Examination of the MMPI-A for the assessm ent of psychopathy in incarcerated male offenders. International Journal of Offender Therapy and Comparative Criminology 403 224-233.

PAGE 242

231 Ireland, T., & Widom, C. S. (1994). Chil dhood victimization and risk for alcohol and drug arrests. The International Jour nal of the Addictions 29 235-274. Ireland, T. O., Smith, C. A., & Thornberry, T. P. (2002). Developmental issues in the impact of child maltreatment on later delinquency and drug use. Criminology 40 (2), 359-396. Jang, S. J., & Johnson, B. R. (2003). St rain, negative emotions, and deviant coping among African Americans: A test of general strain theory. Journal of Quantitative Criminology 19 (1), 79-105. Jensen, G. F. (1995). Salvaging struct ure through strain: A theoretical and empirical critique. In F. Adler & W. S. Laufer (Eds.), The Legacy of Anomie Theory, Advances in Criminological Theory Vol. 6 (pp. 139-158). New Brunswick, NJ: Transaction. Jessor, R., & Jessor, S. L. (1977). Problem Behavior and Psychosocial Development New York: Academic Press. John, O. P., & Srivastava, S. (1999). Th e Big Five trait taxonomy: History, measurement, and theoretical perspectives. In L. A. Pervin & O. P. John (Eds.), Handbook of Personality: Theory and Research (2nd ed., pp. 102-138). New York: The Guilford Press. Johnson, R. A., Hoffman, J. P., Su, S. S., & Gerstein, D. R. (1997). Growth curves of deviant behavior in early adolescence: A multilevel analysis. Journal of Quantitative Criminology 13 (4), 429-467. Johnson, R. E. (1979). Juvenile Delinquency and Its Origins: An Integrated Approach Cambridge, GB: Cambridge University Press. Kagan, J. (1989). Unstable Ideas: Temperament, Cognition, and Self Cambridge, MA: Harvard University Press. Kakar, S. (1996). Child Abuse and Delinquency New York: University Press of America. Kandal, D. (1973). Adolescent marijuana use: Role of parents and peers. Science 181 1067-1070. Kandel, D. B. (1975). Stages in a dolescent involvement in drug use. Science, 190 912-914. Kandel, D. B. (Ed.). (2002). Stages and Pathways of Drug Involvement: Examining the Gateway Hypothesis New York: Cambridge.

PAGE 243

232 Kandel, D. B., Kessler, R. C., & Mar gulies, R. Z. (1978). Antecedents of adolescent initiation into stages of drug use: A developmental analys is. In D. B. Kandel (Ed.), Longitudinal Research on Drug Use (pp. 73-99). Washingt on, DC: Hemisphere. Kandel, D. B., Yamaguchi, K., & Chen, K. (1992). Stages of progression in drug involvement from adolescence to adulthood: Further evidence for the gateway theory. Journal of Studies on Alcohol, 53 447-457. Kaplan, H. B. (1996). Psychological stress fr om the perspective of self theory. In H. B. Kaplan, Psychological Stress San Diego, CA: Academic Press. Karpman, B. (1941). On the need of se parating psychopathy into two distinct clinical types: The symptomatic and the idiopathic. Journal of Criminal Psychopathology, 3 112-137. Karpman, B. (1948). Conscience in the psychopathic personality. American Journal of Psychiatry, 104 523-534. Kessler, R. C., & Greenberg, D. F. (1981). Linear Panel Analysis: Models of Quantitative Change New York: Academic Press. Kim, K. J., Conger, R. D., Elder, G. H., & Lorenz, F. O. (2003). Reciprocal influences between stressful life events a nd adolescent internalizing and externalizing problems. Child Development 74 (1), 127-143. Kochanska, G., Murray, K., & Coy, K. (1997) Inhibitory contro l as a contributor to conscience in childhood: From toddler to early school age. Child Development 68 263-277. Kornhauser, R. R. (1978). Social Sources of Delinquency Chicago: University of Chicago Press. Kosson, D. S., Cyterski, T. D., Steuer wald, B. L., Neumann, C. S., & WalkerMatthews, S. (2002). The reliability and va lidity of the Psychopathy Checklist: Youth Version (PCL: YV) in nonincarcerated adolescent males. Psychological Assessment 14 97-109. Kosson, D. S., Smith, S. S., & Newman, J. P. (1990). Evaluation of the construct validity of psychopathy in black and white male inmates: Three preliminary studies. Journal of Abnormal Psychology 99 250-259. Krohn, M., Akers, R. L., Radosevich, M., & Lanza-Kaduce, L. (1980). Social status and deviance. Criminology, 18 303-318.

PAGE 244

233 Krueger, R. F. (2002). Personality from a realist’s perspective: Personality traits, criminal behaviors, and th e externalizing spectrum. Journal of Research in Personality 36 564-572. Krueger, R. F., Caspi, A., Moffitt, T. E., Silva, P. A., & McGee, R. (1996). Personality traits are differentially linked to mental disorders: A multitrait-multidiagnosis study of an adolescent birth cohort. Journal of Abnormal Psychology 105 299-312. Krueger, R. F., Hicks, B. M., Patrick, C. J., Carlson, S. R., Iacono, W. G., & McGue, M. (2002). Etiological connections among substance dependence, antisocial behavior, and personality: Modeli ng the externalizing spectrum. Journal of Abnormal Psychology 111 (3), 411-424. Krueger, R. F., McGue, M., & Iacono, W. G. (2001). The higher-order structure of common DSM mental disorder s: Internalization, externaliz ation, and their connections to personality. Personality and Individual Differences 30 1245-1259. Krueger, R. F., Schmutte, P. S., Caspi, A., Moffitt, T. E., Campbell, K., & Silva, P. A. (1994). Personality traits are linked to crime among men and women: Evidence from a birth cohort. Journal of Abnormal Psychology 103 328-338. Krueger, R. F., & Tackett, J. L. (2003). Personality and psychopathology: Working toward the bigger picture. Journal of Personality Disorders 17 (2), 109-128. Labouvie, E. W., Pandina, R. J., White, H. R., & Johnson, V. (1990). Risk factors of adolescent drug use: An affect-based interpretation. Journal of Substance Abuse 2 265-285. Lalumiere, M. L., & Quinsey, V. L. ( 1996). Sexual deviance, antisociality, mating effort, and the use of sexually coercive behaviors. Personality and Individual Differences 21 33-48. Lazarus, R. S. (1999). Stress and Emotion: A New Synthesis New York: Springer. Lee, Z., Vincent, G. M., Hart, S. D., & Corrado, R. R. (2003). The validity of the Antisocial Process Screening Device as a self-report measure of psychopathy in adolescents. Behavioral Sciences and the Law 21 771-786. Lemmon, J. H. (1999). How child maltreatme nt affects dimensions of juvenile delinquency in a cohort of low-income urban youths. Justice Quarterly 16 357-376. Lenzenweger, M. F. (1999). Stability and change in personality disorder features. Archives of General Psychiatry 56 (11), 1009-1115.

PAGE 245

234 Lerner, J., & Vicary, J. (1984). Difficult temperament and drug use: Analysis from the New York longitudinal study. Journal of Drug Education 14 1-8. Levenson, M. R., Kiehl, K. A., & Fitzpatrick, C. M. (1995). Assessing psychopathic attributes in a noninstitutional population. Journal of Personality and Social Psychology, 68 151-158. Lilienfeld, S. O. (1999). Anxiety sensitivity and the structure of personality. In S. Taylor (Ed.), Anxiety Sensitivity Mahwah, NJ: Erlbaum. Lilienfeld, S. O., & Andrews, B. P. (1996). Development and preliminary validation of a self-report measure of psyc hopathic personality tr aits in noncriminal populations. Journal of Personality Assessment 66 488-524. Liska, A. E. (1971). Aspirations, expe ctations, and delinquency: Stress and additive models. Sociological Quarterly, 12 99-107. Liska, A. E. (1987). Perspectives on Deviance Englewood Cliffs, NJ: PrenticeHall. Livesley, W. J., Schroeder, M. L., Jackson, D. N., & Jang, K. L. (1994). Categorical distinctions in the study of personality di sorder: Implications for classification. Journal of Abnormal Psychology 103 (1), 6-17. Loeber, R., Farrington, D. P., Stoutha mer-Loeber, M., & Kammen, W. B. van. (1998). Antisocial Behavior and Mental Health Problems: Explanatory Factors in Childhood and Adolescence Mahwah, NJ: Erlbaum. Loney, B. R., Frick, P. J., Clements, C. B., Ellis, M. L., & Kerlin, K. (2003). Callous-unemotional traits, impulsivity, a nd emotional processing in antisocial adolescents. Journal of Clinical Child and Adolescent Psychology 32 66-80. Loukas, A., Krull, J. L., Chassin, L., & Carle, A. C. (2000). The relation of personality to alcohol abuse/depe ndence in a high risk sample. Journal of Personality 68 1153-1175. Luengo, M. A., Otero, J. M., Carrillo-dela-Pea, M. T., & Mirn, L. (1994). Dimensions of antisocial behaviour in j uvenile delinquency: A study of personality variables. Psychology, Crime & Law 1 27-37. Lykken, D. T. (1995). The Antisocial Personalities Hillsdale, NJ: Erlbaum. Lynam, D. R. (1997). Pursuing the psychopath: Capturing the fledgling psychopath in a nomological net. Journal of Abnormal Psychology 106 (3), 425-438.

PAGE 246

235 Lynam, D. R. (2002a). Fledgling psychopa thy: A view from personality theory. Law and Human Behavior 26 (2), 255-259. Lynam, D. R. (2002b). Psychopathy from the perspective of the Five Factor Model. In P. T. Costa & T. A. Widiger (Eds.), Personality Disorders and the Five-Factor Model of Personality (2nd ed.). Washington, D.C.: American Psychological Association. Lynam, D. R., Caspi, A., Moffitt, T. E., Raine, A., Loeber, R., & StouthamerLoeber, M. (in press). Adolescent psychopa thy and the big five: Results from two samples. Journal of Abnormal Child Psychology Lynam, D. R., & Gudonis, L. (in pr ess). The development of psychopathy. Annual Review of Clinical Psychology Lynam, D. R., Leukefeld, C., & Clayt on, R. R. (2003). The contribution of personality to the overlap between antisoc ial behavior and substance use/misuse. Aggressive Behavior 29 316-331. Lynam, D. R., Whiteside, S., & Jones, S. (1999). Self-reported psychopathy: A validation study. Journal of Personality Assessment 73 110-132. Mailloux, D., Forth, A. E., & Kroner, D. G. (1997). Psychopathy and substance abuse in adolescent male offenders. Psychological Reports 80 529-530. Mak, A. S., Heaven, P. C. L., & Rummery, A. (2003). The role of group identity and personality domains as indicat ors of self-reported delinquency. Psychology, Crime & Law 9 9-18. Margolin, G., & Gordis, E. (2000). The eff ects of family and community violence on children. Annual Review of Psychology 51 445-479. Marshall, L., & Cooke, D. (1995). The role of childhood experiences in the aetiology of psychopathy. Issues in Criminological and Legal Psychology 24 107-108. Martin, R. P., Wisenbaker, J., & Huttunen, M. (1994). Review of factor analytic studies of temperament measures base d on the Thomas-Chess structural model: Implications for the Big Five. In C. F. Halv erson, Jr., G. A. Kohnstamm, & R. P. Martin (Eds.), The Developing Structure of Temperam ent and Personality from Infancy to Adulthood (pp. 157-172). Hillsdale, NJ: Erlbaum. Masse, L. C., & Tremblay, R. E. (1997). Be havior of boys in kindergarten and the onset of substance use during adolescence. Archives of General Psychiatry 54 62-68. Matheny, A. P., Jr. (1989). Children’s be havioral inhibition over age and across situations: Genetic similarity for a trait during change. Child Development 54 356-360.

PAGE 247

236 Matza, D. (1964). Delinquency and Drift New York: Wiley. Maxwell, S. R. (2001). A focus on fam ilial strain: Antisocial behavior and delinquency in Filipino society. Sociological Inquiry 71 (3), 265-292. Mazerolle, P. (1998). Gender, general st rain, and delinquency: An empirical examination. Justice Quarterly 15 (1), 65-91. Mazerolle, P., Burton, V. S., Cullen, F. T., Evans, T. D., & Payne, G. L. (2000). Strain, anger, and delinquent adaptati ons specifying genera l strain theory. Journal of Criminal Justice 28 89-101. Mazerolle, P., & Maahs, J. (2000). General strain and delinquency: An alternative examination of conditioning influences. Justice Quarterly 17 (4), 753-778. Mazerolle, P., & Piquero, A. (1997). Violent responses to situations of strain: A structural examination of conditioning effects. Violence and Victims 12 323-344. Mazerolle, P., & Piquero, A. (1998). Linking exposure to strain with error: An investigation devi ant adaptations. Journal of Criminal Justice 26 (3), 195-211. Mazerolle, P., Piquero, A., & Capowich, G. E. (2003). Examining the links between strain, situational and dispositional anger, and crime: Further specifying and testing general strain theory. Youth & Society 35 (2), 131-157. McCoy, M. G., Frick, P. J., Loney, B. R., & Ellis, M. L. (1999). The potential mediating role of parenting practices in the development of conduct problems in a clinicreferred sample. Journal of Children and Family Studies, 8 477-494. McCrae, R. R. (1994). New goals for trait psychology. Psychological Inquiry 5 148-153. McCrae, R. R., & Costa, P. T. (1990). Personality in Adulthood New York: Guilford Press. McCrae, R. R., & Costa, P. T. (1996). Toward a new generation of personality theories: Theoretical contexts for the five -factor model. In J. S. Wiggins (Ed.), The FiveFactor Model of Personality: Theoretical Perspectives (pp. 51-87). New York: Guilford Press. McCrae, R. R., & John, O. P. (1992). An in troduction to the five-factor model and its application. Journal of Personality 60 175-216.

PAGE 248

237 McDevitt, S. C. (1986). Continuity and di scontinuity of temperament in infancy and early childhood: A psychometric perspec tive. In R. Plomin & J. Dunn (Eds.), The Study of Temperament: Changes, Continuities and Challenges (pp. 27-38). Hillsdale, NJ: Erlbaum. McGue, M., Bacon, S., & Lykken, D. T. ( 1993). Personality stability and change in early adulthood: A beha vioral genetic analysis. Developmental Psychology 29 96109. McGue, M., Slutske, W., & Iacono, W. G. (1999). Personality and substance use disorders II: Alcoholism versus drug use disorders. Journal of Consulting and Clinical Psychology 67 394-404. McNally, S., Eisenberg, S., & Harris, J. (1991). Consistency and change in maternal child-rearing practices and values: A longitudinal study. Child Development 62 190-198. Mealey, L. (1995a). Primary sociopathy (psy chopathy) is a type, secondary is not. Behavioral and Brain Sciences 19 579-599. Mealey, L. (1995b). The sociobiology of so ciopathy: An integrated evolutionary model. Behavioral and Brain Sciences 19 523-540. Mednick, S. A., Gabrielli, W., & Hutchi ngs, B. (1984). Genetic influences in criminal convictions: Evidence from an adoption cohort. Science 224 891-893. Menard, S. (1995). A developmental test of Mertonian anomie theory. Journal of Research in Crime and Delinquency, 32 (2), 136-174. Merton, R. K. (1938). Social structure and anomie. American Sociological Review, 3 672-682. Merton, R. K. (1957). Social Theory and Social Structure (Rev. ed.). Glencoe, IL: Free Press. Merton, R. K. (1968). Social Theory and Social Structure New York: Free Press. Messner, S. (1985). Sex differen ces in arrest rates for ho micide: An application of general theory of structural strain. Comparative Sociological Research 8 187-201. Messner, S. (1988). Merton’s “social struct ure and anomie”: The road not taken. Deviant Behavior 9 33-53. Messner, S. F., & Rosenfeld, R. (1994). Crime and the American Dream Belmont, CA: Wadsworth.

PAGE 249

238 Messner, S. F., & Rosenfeld, R. (1997). Politic al restraint of the market and levels of criminal homicide: A cross-national application of institutional anomie. Social Forces, 75 1393-1416. Meyers, K., Hagan, T. A., McDermott, P., Webb, A., Randall, M., & Frantz, J. (in press). Design and validation of clinical di mensions for the Comprehensive Adolescent Severity Inventory (CASI). The American Journal of Drug and Alcohol Abuse Meyers, K., Hagan, T. A., Zanis, D., Webb, A., Frantz, J., Ring-Kurtz, S., Rutherford, M., & McLellan, A. T. (1999). Crit ical issues in adol escent substance use assessment. Drug and Alcohol Dependence 55 235-246. Meyers, K., Webb, A., Hagan, T. A., & Frantz. (submitted). Do adolescents reliably report age of problem onset a nd history? An inve stigation using the Comprehensive Adolescent Se verity Inventory (CASI). Miller, J. D., & Lynam, D. (2001). Struct ural models of personality and their relation to antisocial behavior: A meta-analytic review. Criminology 39 (4), 765-798. Miller, J. D., Lynam, D., R., Widiger, T. A., & Leukefeld, C. (2001). Personality disorders as extreme variants of common pe rsonality dimensions: Can the Five-Factor Model adequately represent psychopathy? Journal of Personality 69 253-276. Mischel, W. (1968). Personality and Assessment New York: Wiley. Moffitt, T. E. (1987). Parental mental di sorder and offspring criminal behavior: An adoption study. Psychiatry 50 346-360. Moffitt, T. E. (1993). Adolescent-limited and life-course-persistent antisocial behaviour: A developmental taxonomy. Psychological Review 100 674-701. Molt, J., Poy, R., & Torrubia, R. (2 000). Standardization of the Hare Psychopathy Checklist-Revised in a Spanish prison sample. Journal of Personality Disorder 14 84-96. Moss, H. B., Lynch, K. G., & Hardie, T. L. (2003). Affiliation with deviant peers among children of substance dependent father s from pre-adolescence into adolescence: Associations with problem behaviors. Drug and Alcohol Dependence, 71 (2), 117-125. Murrie, D., & Cornell, D. (2002). Psychopa thy screening of incarcerated juvenile: A comparison of measures. Psychological Assessment 14 390-396. Murrie, D. C., Cornell, D. G., Kapla n, S., McConville, D., & Levy-Elkon, A. (2004). Psychopathy scores and violence am ong juvenile offenders: A multi-measure study. Behavioral Sciences and the Law 22 49-67.

PAGE 250

239 Muthn, L. K., & Muthn, B. O. (2004). Mplus Users’ Guide (3rd ed.). Los Angeles, CA: Muthn & Muthn. Myers, W. C., Burket, R. C., & Harris, H. E. (1995). Adolescent psychopathy in relation to delinquent behaviors, conduct disorders, and personality disorders. Journal of Forensic Science 40 436-440. Nagin, D. S., & Farrington, D. P. (1992). Th e stability of criminal potential from childhood to adulthood. Criminology 30 235-260. Nagin, D. S., & Paternoster, R. (1991). On the relationship past to future participation in delinquency. Criminology 29 163-189. Newcomb, M. D., Chou, C., Bentler, P. M., & Huba, G. J. (1988). Cognitive motivations for drug use among adolescents: Lo ngitudinal tests of ge nder differences and predictors of change in drug use. Journal of Counseling Psychology 35 426-438. Newcomb, M. D., & McGee, L. (1991). Infl uence of sensation seeking on general deviance and specific problem behavior s from adolescence to young adulthood. Journal of Personality and Social Psychology 61 614-628. Nye, F. I. (1958). Family Relationships and Delinquent Behavior New York: Wiley. O’Brien, B. S., & Frick, P. J. (1996). Reward dominance: Associations with anxiety, conduct problems, and psychopathy in children. Journal of Abnormal Child Psychology 24 223-240. O’Neill, M. L., Lidz, V., & Heilbrun, K. (2003). Adolescents with psychopathic characteristics in a substance abusing cohort: Treatment process and outcomes. Law and Human Behavior 27 299-313. Ogloff, J.R.P., Wong, S., & Greenw ood, A. (1990). Treating criminal psychopaths in a therapeutic community program. Behavioral Sciences and the Law 8 181-190. Osgood, D. W., Johnston, L. D., O’Malley, P. M., & Bachman, J. G. (1988). The generality of deviance in late adolescence and early adulthood. American Sociological Review 53 81-93. Ostendorf, F., & Angleitner, A. (199 4). Enthusiasts contra pessimists. Psychological Inquiry 5 159-162.

PAGE 251

240 Pals, J. L. (1999). Identity consolidati on in early adulthood: Relations with egoresiliency, the context of marri age, and personality change. Journal of Personality 67 295-329. Pardini, D., Lochman, J., & Frick, P. (2003) Callous/unemotional traits and social cognitive processes in adjudicated youth. Journal of American Academy of Child and Adolescent Psychiatry 19 347. Passas, N. (1995). Continuities in the anomie tradition. In F. Adler & W. S. Laufer (Eds.), The Legacy of Anomie Theory, Ad vances in Criminological Theory Vol. 6 (pp. 91-112). New Brunswick, NJ: Transaction. Paternoster, R., & Mazerolle, P. (1994). General strain theory and delinquency: A replication and extension. Journal of Research in Crime and Delinquency 31 (3), 235263. Patrick, C. J. (1994). Emotion and psychopathy. Psychophysiology 31 319-330. Patrick, C. J., Bradley, M. M., & Lang, P. J. (1993). Emotion in the criminal psychopath: Startle reflex modulation. Journal of Abnormal Psychology 102 82-92. Patrick, C. J., Edens, J. F., Poythress, N., Lilienfeld, S. O., & Benning, S. D. (manuscript in preparation). Further Evidence for Two Distinct Facets of Self-Reported Psychopathy: Differential Relations with Criterion Variables in Offender Samples Manuscript in prepar ation for publication. Patterson, G. R. (1995). Coercion as a basis fo r early age of onset for arrest. In J. McCord (Ed.), Coercion and Punishment in Long-Term Perspectives (pp. 81-105). Cambridge: Cambridge University Press. Patterson, G. R., & Chamberlain, P. (1994) A functional analysis of resistance during parent training therapy. Clinical Psychology: Science and Practice 1 53-70. Patterson, G. R., Reid, J. B., & Dishion, T. J. (1992). Antisocial Boys Eugene, OR: Castalia Publishing Co. Patterson, G. R., Reid, J. B., Jones, R. Q., & Conger, R. E. (1975). A Social Learning Approach to Family Intervention Vol. 1. Eugene, OR: Castalia Publishing Co. Patterson, M. C., & Newman, J. P. (1993). Reflectivity and lear ning from aversive events: Toward a psychological mechan ism for the syndromes of disinhibition. Psychological Review, 100 716-736. Pedhazur, E. J., & Schmelkin, L. P. (1991). Measurement, Design, and Analysis: An Integrated Approach Hillsdale, NJ: Lawrence Erlbaum Associates.

PAGE 252

241 Pedlow, R., Sanson, A., Prior, M., & Oberkl aid, F. (1993). Stability of maternally reported temperament from infancy to 8 years. Developmental Psychology 29 998-1007. Peter, T., LaGrange, T. C., & Silver man, R. A. (2003). Investigating the interdependence of strain and self-control. Canadian Journal of Criminology and Criminal Justice, 45 (4), 431-464. Piatigorsky, A., & Hinshaw, S. P. (2004) Psychopathic traits in boys with and without Attention-Deficit/Hypera ctivity Disorder: Concurrent and longitudinal correlates. Journal of Abnormal Child Psychology 32 (5), 535-550. Piliavin, I., Thornton, C., Gartner, R., & Matsueda, R. L. (1986). Crime, deterrence, and rational choice. American Sociological Review 51 101-119. Piquero, A., & Piquero, N. L. (1998). On te sting institutional anomie theory with varying specifications. Studies on Crime and Crime Prevention, 7 61-84. Piquero, N. L., Gover, A. R., MacDonal d, J. M., & Piquero, A. R. (2005). The influence of delinquent peers on delinquency: Does gender matter? Youth & Society, 36 (3), 251-275. Piquero, N. L., & Sealock, M. D. (2000). Generalizing general strain theory: An examination of an offending population. Justice Quarterly, 17 449-484. Piquero, N. L., & Sealock, M. D. (2004) Gender and general strain theory: A preliminary test of Broidy a nd Agnew’s gender/GST hypotheses. Justice Quarterly 21 (1), 125-158. Plomin, R., & Nesselrode, J. (1990). Beha viour genetics and personality change. Journal of Personality 50 191-219. Porter, S. (1996). Without conscience or without active cons cience? The etiology of psychopathy revisited. Aggression and Violent Behavior 1 179-189. Poythress, N., Dembo, R., & DuDell, G. (2004). Arbitration Intervention Worker (AIW) Services: Case Management Overlay in a Juvenile Diversion Program Unpublished manuscript, University of South Florida at Tampa. Poythress, N., Dembo, R., Wareham, J., & Greenbaum, P. E. (in press). Construct validity of the Youth Psychopathic features Inventory (YPI) and th e Antisocial Process Screening Device (APSD) with ju stice-involved adolescents. Criminal Justice and Behavior Pratt, T. C., & Godsey, T. W. (2003). So cial support, inequality, and homicide: A cross-national test of an integrated theoretical model. Criminology, 41 (3), 611-643.

PAGE 253

242 Pulkkinen, L. (1983). Youthful smok ing and drinking in a longitudinal perspective. Journal of Youth and Adolescence 12 253-283. Quay, H. C. (1987). Patterns of delinque nt behavior. In H. C. Quay (Ed.), Handbook of Juvenile Delinquency New York: Wiley. Quicker, J. C. (1974). The effect of goal discrepancy on delinquency. Social Problems, 22 76-86. Raine, A. (1993). The Psychopathology of Crime: Criminal Behavior as a Clinical Disorder San Diego, CA: Academic Press. Read, S. J., Jones, D. K., & Miller, L. C. (1990). Traits as goal-based categories: The importance of goals in the coherence of dispositional categories. Journal of Personality and Social Psychology 58 1048-1061. Reiss, A. J. (1951). Delinquency as the fa ilure of personal and social controls. American Sociological Review 16 196-207. Reiss, D., Grubin, D., & Meux, C. (1999) Institutional perf ormance of male “psychopaths” in a high-security hospital. Journal of Forensic Psychiatry 10 290-299. Rice, M. E., & Harris, G. T. (1995). Ps ychopathy, schizophrenia, alcohol abuse, and violent recidivism. International Journal of Law and Psychiatry 18 333-342. Rice, M. E., Harris, G. T., & Cormier, C. A. (1992). An evaluation of a maximum security therapeutic community for psychopaths and other mentally disordered offenders. Law and Human Behavior 16 399-412. Rice, M. E., Harris, G. T., & Quinsey, V. L. (1990). A follow-up of rapists assessed in a maximum securi ty psychiatric facility. Journal of Interp ersonal Violence 5 435-448. Ridenour, T. A. (2001). Is the revised Ps ychopathy Checklist clin ically useful for adolescents? Journal of Psychoeducational Assessment 19 227-238. Robbers, M. L. P. (2004). Revisiting the moderating effect of social support on strain: A gendered test. Sociological Inquiry 74 (4), 546-596. Roberts, B. W., & DelVecchio, W. F. (2000). The rank-order consistency of personality from childhood to old age: A qua ntitative review of l ongitudinal studies. Psychological Bulletin 126 3-25. Roberts, G. C., Block, J. H., & Block, J. (1984). Continuity and change in parents’ child rearing practices. Child Development 55 586-597.

PAGE 254

243 Robins, L. N. (1966). Deviant Children Grow Up: A Sociological and Psychiatric Study of Sociopathic Personality Baltimore, MD: Williams & Wilkins. Rogers, R., Johansen, J., Chang, J. J., & Salekin, R. T. (1997). Predictors of adolescent psychopathy: Oppositional and conduct-disordered symptoms. Journal of the American Academy of Psychiatry and Law 25 261-271. Romero, E., Luengo, M. A., & Sobral, J. (2001). Personality and antisocial behaviour: Study of temperamental dimensions. Personality and Individual Differences 31 329-348. Rothbart, M. K., & Bates, J. E. (199 8). Temperament. In W. Damon & N. Eisenberg (Eds.), Handbook of Child Psychology: Vol. 3. Social, Emotional, and Personality Development (5th ed., pp. 311-388). New York: Wiley. Rotter, J. (1954). Social Learning and Clinical Psychology Englewood Cliffs, NJ: Prentice-Hall. Royce, D., & Wiehe, V. R. (1988). Impul sivity in felons and unwed mothers. Psychological Reports 62 335-336. Rutherford, M. J., Alterman, A. I., C acciola, J. S., & McKay, J. R. (1997). Validity of the Psychopathy Checklist-Rev ised in male methadone patients. Drug and Alcohol Dependence 44 143-149. Rutherford, M. J., Cacciola, J. S., Alterman, A. I., & McKay, J. R. (1996). Reliability and validity of the Psychopat hy Checklist in women methadone patients. Assessment 3 145-156. Rutter, M. (1996). Introduction: Concepts of antisocial behavior, of cause and of genetic influences. In G. R. Bock & J. A. Goode (Eds.), Genetics of Criminal and Antisocial Behavior Toronto: Wiley. Rutter, M., & Giller, H. (1983). Juvenile Delinquency: Trends and Perspectives Harmondsworth: Penguin. Salekin, R. T. (2002). Psychopathy and ther apeutic pessimism: Clinical lore or clinical reality? Clinical Psychology Review 22 79-112. Salekin, R. T., Leistico, A. R., Neumann, C. S., DiCicco, T. M., & Duros, R. L. (2004). Psychopathy and comorbidity in a young offender sample: Taking a closer look at psychopathy’s potential importance over disruptive behavior disorders. Journal of Abnormal Psychology 113 416-427.

PAGE 255

244 Salekin, R. T., Leistico, A. R., Trobst, K. K., Schrum, C. L., & Lochman, J. E. (in press). Adolescent psychopathy and the interp ersonal circumplex: E xpanding evidence of a nomological net. Journal of Abnormal Child Psychology Salekin, R. T., Rogers, R., & Sewell, K. W. (1996). A review and meta-analysis of the Psychopathy Checklist and Psychopathy Checklist-Revised: Pr edictive validity of dangerousness. Clinical Psychology: Science and Practice 3 625-637. Salekin, R. T., Rogers, R., & Sewell, K. W. (1997). Construct validity of psychopathy in a female offender sample: A multitrait-multimethod evaluation. Journal of Abnormal Psychology 106 576-585. Salekin, R. T., Rogers, R., Ustad, K. L ., & Sewell, K. W. (1998). Psychopathy and recidivism among female inmates. Law and Human Behavior 22 109-128. Sampson, R. J., & Laub, J. H. (1993). Crime in the Making: Pathways and Turning Points Through Life Cambridge, MA: Harvard University Press. Sampson, R. J., & Laub, J. H. (1997). A life course theory of cumulative disadvantage and the stability of deli nquency. In T. P. Thornberry (Ed.), Developmental Theories of Crime and Delinquency New Brunswick, NJ: Transaction Publishers. Sampson, R. J., & Laub, J. H. (2003). Life-c ourse desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology 41 (3), 555-592. Savolainen, J. (2000). Inequali ty, welfare state, and homi cide: Further support for the institutional anomie theory. Criminology, 38 1021-1042. Sayette, M. A. (1993). An appraisal-disruption model of alcohol’s effects on stress responses in social drinkers. Psychological Bulletin 114 459-476. Schmitt, W. A., & Newman, J. P. (1999) Are all psychopathic individuals lowanxious? Journal of Abnormal Psychology 108 353-358. Schneider, K. (1958). Psychopathic Personalities (9th ed., M. Hamilton, Trans.). London: Cassell. (Original work published 1950) Schuerger, J. M., Zarrella, K. L., & Hotz, A. S. (1989). Factors that influence the temporal stability of pe rsonality by questionnaire. Journal of Personality and Social Psychology 56 777-783. Schuessler, K. F., & Cressey, D. F. ( 1950). Personality characteristics of criminals. American Journal of Sociology 55 476-484.

PAGE 256

245 Seagrave, D., & Grisso, T. (2002). Adol escent development and the measurement of juvenile psychopathy. Law and Human Behavior 26 219-239. Segrave, J. O., & Halsted, D. N. (1983). Evaluating structural and control models of delinquency causation: A replication. Youth and Society, 14 437-456. Serin, R. C. (1991). Psychopathy and violence in criminals. Journal of Interpersonal Violence 6 423-431. Serin, R. C., Peters, R. D., & Barbaree, H. E. (1990). Predictors of psychopathy and release outcome in a criminal population. Psychological Assessment 2 419-422. Seto, M. C., & Barbaree, H. E. (1999). Psychopathy, treatment behavior, and sex offender recidivism. Journal of Interpersonal Violence 14 1235-1248. Sharp, S. F., Brewster, D., & Love, S. R. (2005). Disentangling strain, personal attributes, affective response a nd deviance: A gendered analysis. Deviant Behavior, 26 133-157. Sharp, S. F., Terling-Watt, T. L., Atkins, L. A., & Gilliam, J. T. (2001). Purging behavior in a sample of colle ge females: A research note on general strain theory and female deviance. Deviant Behavior: An In terdisciplinary Journal 22 171-188. Shaw, C., & McKay, H. D. (1942). Juvenile Delinquency and Urban Areas Chicago: University of Chicago Press. Shoal, G. D., & Giancola, P. R. (2001). Cognition, negative affectivity, and substance abuse in adolescent boys with and wi thout a family history of a substance use disorder. Journal of Studies on Alcohol 62 675-686. Shoal, G. D., & Giancola, P. R. (2003). Negative affectivity and drug use in adolescent boys: Moderating and mediating mechanisms. Journal of Personality and Social Psychology 84 (1), 221-233. Sigfusdottir, I.-D., Farkas, G., & Silver, E. (2004). Th e role of depressed mood and anger in the relationsh ip between family conflic t and delinquent behavior. Journal of Youth and Adolescence 33 (6), 509-522. Silva, F., Martorell, M. C., & Clemente, A. (1986). Socialization and personality: Study through questionnaires in a preadult Spanish population. Personality and Individual Differences 7 355-372. Sim, S., & Perez, J. (1991). Sensation s eeking and antisocial be havior in a junior high school sample. Personality and Individual Differences 12 965-966.

PAGE 257

246 Simons, R. L., Wu, C.-I., Conger, R. D., & Lorenz, F. O. (1994). Two routes to delinquency: Differences between early and late starters in the impact of parenting and deviant peers. Criminology 32 (2), 247-276. Sirles, E. A. (1990). Dropout from in take, diagnostics, and treatment. Community Mental Health Journal, 26 345-360. Skeem, J. L., & Cauffman, E. (2003). Views of the downward extension: Comparing the youth version of the Psychopath ic Checklist tithe the Youth Psychopathic traits Inventory. Behavioral Sciences and the Law 21 737-770. Skeem, J. L., Edens, J. F., Camp, J., & Colwell, L. H. (2004). Are there ethnic differences in levels of ps ychopathy? A meta-analysis. Law and Human Behavior 28 (5), 505-527. Skeem, J. L., Poythress, N., Edens, J. F ., Lilienfeld, S. O., & Cale, E. M. (2003). Psychopathic personality or personalities? Ex ploring potential vari ants of psychopathy and their implications for risk assessment. Aggression and Violent Behavior 8 513-546. Smith, C., & Thornberry, T. P. (1995) The relationship between childhood maltreatment and adolescent involvement in delinquency. Criminology 33 451-477. Smith, P. K., & Myron-Wilson, R. (1998) Parenting and school bullying. Clinical Child Psychology and Psychiatry 3 405-417. Smith, S. S., & Newman, J. P. (1990) Alcohol and drug abuse-dependence disorders in psychopathic and nonps ychopathic criminal offenders. Journal of Abnormal Psychology 4 430-439. Soldz, S., & Vaillant, G. E. (1999). The Bi g Five personality traits and the life course: A 45-year longitudinal study. Journal of Research in Personality 33 208-232. Spain, S. E., Douglas, K. S., Poythres s, N. G., & Epstein, M. (2004). The relationship between psychopathic featur es, violence and treatment outcome: The comparison of three youth measur es of psychopathic features. Behavioral Sciences and the Law 22 85-102. Stafford, E., & Cornell, D. G. (2003). Psychopathy scores predict adolescent inpatient aggression. Assessment 10 102-112. Stein, J. A., Newcomb, M. D., & Bentler, P. M. (1986). Stability and change in personality: A longitudinal study from early adolescence to young adulthood. Journal of Research in Personality 20, 276-291.

PAGE 258

247 Steinberg, L. (2001). The juvenile psychopa th: Fads, fictions, and facts. National Institute of Justice Perspec tives on Crime and Justice: 2001 Lecture Series, vol. V, 35-64. Stevens, D. P., & Truss, C. V. (1987). St ability in adult personality over 12 and 20 years. Developmental Psychology 21 568-584. Stewart, S. H., Karp, J., Pihl, R. O., & Peterson, R. A. (1997). Anxiety sensitivity and self-reported r easons for drug use. Journal of Substance Abuse 9 223-240. Stice, E., & Gonzales, N. (1998). Adoles cent temperament mode rates the relation of parenting to antisocial behavior and substance use. Journal of Adolescent Research 13 5-31. Stormshak, E. A., Bierman, K. L., McMa hon, R. J., & Lengua, L. J. (2000). Parenting practices and child disruptive behavior in early elementary school. Journal of Clinical Child Psychology 29 17-29. Sutherland, E. H. (1947). Principles of Criminology (4th ed.). Philadelphia: J. B. Lippincott. Tarter, R. E., Blackson, T., Brigham, J., Moss, H., & Caprara, G. V. (1995). The association between childhood ir ritability and liability to substance use in early adolescence: A 2-year follow-up study of boys at risk for substance abuse. Drug and Alcohol Dependence 39 253-261. Tedeschi, J. T., & Felson, R. B. (1994). Violence, Aggression, and Coercive Actions Washington, DC: American Psychological Association. Tedeschi, J. T., & Nesler, M. S. (1993) Grievances: Development and reaction. In R. B. Felson & J. T. Tedeschi (Eds.), Aggression and Violence Washington, DC: American Psychological Association. Tellegen, A. (1985). Structures of mood a nd personality and their relevance to assessing anxiety with emphasi s on self-report. In A. H. Tuma & J. D. Maser (Eds.), Anxiety and the Anxiety Disorders Hillsdale, NJ: Lawrence Erlbaum Associates. Tennenbaum, D. J. (1977). Personality and criminality: A summary and implications of the literature. Journal of Criminal Justice 5 225-235. Thaxton, S., & Agnew, R. (2004). The nonlin ear effects of parental and teacher attachment on delinquency: Disentangling st rain from social c ontrol explanations. Justice Quarterly, 21 (4), 763-791. Thomas, A., & Chess, S. (1977). Temperament and Development New York: Brunner/Mazel.

PAGE 259

248 Thornberry, T., & Farnworth, M. (1982) Social correlates of criminal involvement: Further evidence on the relations hip between social st atus and criminal behavior. American Sociological Review, 47 505-518. Tolan, P., Ryan, K., & Jaffe, C. (1988). A dolescents’ mental health service use and provider, process, and recipient characteristics. Journal of Clinical Child Psychology, 17 229-236. Toupin, J., Mercier, H., Dery, M., Cote, G., & Hodgins, S. (1 995). Validity of the PCL-R for adolescents. Issues in Criminological and Legal Psychology 24 143-145. Tourian, K., Alterman, A., Metzger, D ., Rutherford, M., Cacciola, J. S., & McKay, J. R. (1997). Validity of three measur es of antisociality in predicting HIV risk behaviors in methadone-maintenance patients. Drug and Alcohol Dependence 47 99107. Tremblay, R. E., Pihl, R. O., Vitaro, F ., & Dobkin, P. L. (1994). Predicting early onset of male antisocial beha vior from preschool behavior. Archives of General Psychiatry 51 732-739. Tucker, L. R., & Lewis, C. (1973). A reliability coefficient for maximum likelihood factor analysis. Psychometrika, 38 1-10. Turner, R. J., Wheaton, B., & Lloyd, D. A. (1995). The epidemiology of social stress. American Sociological Review 60 104-125. Tyler, T. R., Boeckmann, R. J., Smith, H. J., & Huo, Y. J. (1997). Social Justice in a Diverse Society Boulder, CO: Westview. Uehara, T., Sakado, K., Sakado, M., Sato, T., & Soomeya, T. (1999). Relationships between stress and coping a nd personality in patients with major depressive disorder. Psychotherapy and Psychosomatics 68 26-30. Verona, E., Patrick, C. J., & Joiner T. T. (2001). Psychopathy, antisocial personality, and suicide risk. Journal of Abnormal Psychology 110 462-470. Vitacco, M. J., Rogers, R., & Neumann, C. S. (2003). The Antisocial Process Screening Device: An examination of its construct and criterion-related validity. Assessment 10 143-150. Vitale, J. E., Smith, S. S., Brinkle y, C. A., & Newman, J. P. (2002). The reliability and validity of the Psychopathy Checklist—Revised in a sample of female offenders. Criminal Justice and Behavior 29 202-231.

PAGE 260

249 Voss, H. L. (1966). Socio-economic stat us and reported delinquent behavior. American Journal of Community Psychology, 11 169-183. Wachs, T. D. (1994). Fit, context, an d the transition between temperament and personality. In C. F. Halverson, Jr., G. A. Kohnstamm, & R. P. Martin (Eds.), The Developing Structure of Temperament and Personality from Infancy to Adulthood (pp. 209-220). Hillsdale, NJ: Erlbaum. Waldo, G. P., & Dinitz, S. (1967). Persona lity attributes of the criminal: An analysis of resear ch studies, 1950-1965. Journal of Research on Crime and Delinquency 4 185-202. Wallace, L. H., Patchin, J. W., & May, J. D. (2005). Reactions of victimized youth: Strain as an explanation of school delinquency. Western Criminology Revie w, 6 (1), 104-116. Walsh, A. (2000). Behavioral gene tics and anomie/strain theory. Criminology 38 1075-1108. Warner, B. D., & Fowler, S. K. (2003). Strain and violence: Testing a general strain theory model of community violence. Journal of Criminal Justice 31 511-521. Warr, M. (2002). Companions in Crime New York: Cambridge University Press. Watson, D., Clark, L. A., & Harkness, A. (1994). Structures of personality and their relevance to psychopathology. Journal of Abnormal Psychology 103 18-31. Weiler, B., & Widom, C. (1996). Psychopat hy and violent behavior in abused and neglected young adults. Criminal Behavior and Mental Health, 6 253-271. Weinrich, S., Hardin, S., Weinrich, M., & Valois, R. F. (1997). Social support and psychological correlates of high school stude nts who use illicit dr ugs in response to stress. American Journal of Health Studies 13 17-26. Wheaton, B. (1994). Sampling the stress univer se. In W. R. Avison & I. H. Gotlib (Eds.), Stress and Mental Health: Contemporar y Issues and Prospects for the Future (pp. 77-114). New York: Plenum Press. White, J. E., Moffitt, T. E., Caspi, A., Bartusch, D. J., Needles, D. J., & Stouthamer-Loeber, M. (1994). Measuring impul sivity and examining its relationship to delinquency. Journal of Abnormal Psychology 103 192-205. Widiger, T. A., & Lynam, D. R. (1998) Psychopathy and th e Five-Factor Model of personality. In T. Millon, E. Simonsen, M. Birket-Smith, & R. D. Davis (Eds.), Psychopathy: Antisocial, Criminal, and Violent Behaviors New York: Guilford.

PAGE 261

250 Widom, C. S. (1977). A methodolog y for studying noninstitutionalized psychopaths. Journal of Consulting and Clinical Psychology 45 674-683. Widom, C. S. (1991). Childhood victimiza tion: Risk factors for delinquency. In M. E. Colton & S. Gore (Eds.), Adolescent Stress: Causes and Consequences New York: Aldine de Gruyter. Widom, C. S., & Newman, J. P. (1985). Characteristics of noninstitutionalized psychopaths. In D. R. Farrington & J. Gunn (Eds.), Aggression and Dangerousness New York: Wiley. Wiggins, J. S. (Ed.). (1996). The Five-Factor Model of Personality: Theoretical Perspectives New York: Guilford Press. Williamson, S., Hare, R. D., & Wong, S. (1987). Violence: Criminal psychopaths and their victims. Canadian Journal of Behavioural Sciences 19 (4), 454-462. Wills, T. A., Sandy, J. M., Shinar, O., & Yaeger, A. (1999). Contributions of positive and negative affect to adolescent substance use: Test of a bidimensional model in a longitudinal study. Psychology of Addictive Behaviors 13 327-338. Wilson, J. J., Rojas, N., Haapanen, R ., Duxbury, E., & Steiner, H. (2001). Substance abuse and criminal recidivism : A prospective study of adolescents. Child Psychiatry and Human Development 31 (4), 297-312. Windle, M. (1990). A longitudinal study of antisocial behaviors in early adolescence as predictors of late adolescent substance use: Gender and ethnic group differences. Journal of Abnormal Psychology 99 86-91. Windle, M. (1991). The difficult temperamen t in adolescence: Association with substance use, family support, and problem behaviors. Journal of Clinical Psychology 47 310-315. Winter, D. G. (1996). Personality: Analysis and Interpretation of Lives New York: McGraw-Hill. Winter, D. G., John, O. P., Stewart, A. J., Klohnen, E. C., & Duncan, L. E. (1998). Traits and motives: Toward an integr ation of two traditions in personality research. Psychological Review 105 (2), 230-250. Wofford, J. C., Daly, P. S., & Juban, R. L. (1999). Cognitive processes, strain, and stress: Gender differences in stress propensity indicators. Anxiety, Stress, and Coping 12 41-62.

PAGE 262

251 Wolfgang, M. E., Figlio, R. M., & Sellin, T. (1972). Delinquency in a Birth Cohort Chicago: University of Chicago Press. Wong, S. (1984). The Criminal and Institutional Behaviours of Psychopaths Ottawa: Ministry of Solicitor General. Wooton, J. M., Frick, P. J., Shelton, K. K., & Silverhorn, P. (1997). Ineffective parenting and childhood conduct problems: The moderating role of callous-unemotional traits. Journal of Consulting and Clinical Psychology 65 301-308. Zillman, D. (1979). Hostility and Aggression Hillsdale, NJ: Lawrence Erlbaum. Zuckerman, M. (1979). Sensation Seeking: Beyond the Optimal Level of Arousal Hillsdale, NJ: Erlbaum. Zuckerman, M. (1989). Personality in th e third dimension: A psychobiological approach. Personality and Individual Differences 10 391-418. Zuckerman, M. (1994). Behavioral Expressions and Bios ocial Bases of Sensation Seeking Cambridge: Cambridge University Press. Zweig, D., & Webster, J. (2004). What are we measuring? An examination of the relationships between the big-five personali ty traits, goal orientation, and performance intentions. Personality and Individual Differences 36 1693-1708.

PAGE 263

252 Appendices

PAGE 264

253 Appendix A: Varimax Rotated Exploratory Fact or Analyses Results for Family, Peer, and School Items during Twelve Months Pr ior to Baseline Interview (N = 137) Family Items Factor 1 Factor 2 Factor 3 1. Repeatedly insulted/criticized .72 .34 .37 2. Other member insulted criticized .70 -.18 .54 3. Other member threw object, punched walls .02 .20 .68 4. Hit hard (physically abused) -.01 -.03 .98 5. Couldn’t get along/fighting with family member .10 .50 .26 6. Ignored or given “silent treatment” .79 .39 .18 7. Other member ignored or “silent treatment” .11 .33 .63 8. Family contacted about domestic disputes .31 .27 .18 9. Parents disagree on limits/punishment .29 .67 .15 10. Home felt like safe place [reversed] .52 .32 -.24 11. Family works out problems non-violently [reversed] .61 .28 .05 12. Hard to talk to/confide in parents .21 .76 .13 13. Ran away from home .46 .42 -.17 14. Parents don’t listen to you .26 .78 .17 15. Parents unavailable to you .44 .61 -.09 16. Parents covered/made excuses for you .21 .37 .09 17. Rules not consistently enforced .08 .62 -.04 18. Felt loved by someone in home [reversed] .76 .22 -.43 19. Given praise for good behavior [reversed] .41 .53 -.09 20. Parents really know what/where you go/do [reversed] .16 .37 -.40 Eigenvalue = 3.81 4.12 2.86 Variance = 19.1 20.6 14.3 Promax factor correlations: 1 with 2 = .525; 1 with 3 = -.007; 2 with 3 = .017 Chi-square = 38.86, df = 29, p = 0.10; RMSEA = 0.050

PAGE 265

254 Appendix A: (Continued) Peer Items Factor 1 Factor 2 1. Difficulty making/keeping friends .90 .03 2. Had no friends .58 .05 3. Preferred to be alone .57 .33 4. Felt friends were not loyal .74 .17 5. Hard to talk to friends .82 .27 6. Dissatisfied with quality of friendships .80 .18 7. Consistently teased/bullied .55 -.06 8. Hung out with people who use drugs/drink .06 .93 9. Hung out with people who commit illegal acts .00 .78 10. Hung out with gang members .19 .76 11. Hung out with people who skipped/dropped school .17 .66 Eigenvalue = 3.68 2.71 Variance = 33.4 24.6 Promax factor correlations: 1 with 2 = .296 Chi-square = 22.08, df = 17, p = 0.18; RMSEA = 0.047

PAGE 266

255 Appendix A: (Continued) School Items Factor 1 Factor 2 1. Had failing grades/difficulty learning .74 .16 2. Skipped class/arrived late consistently .56 .19 3. Went to school prepared [reversed] .36 .62 4. Felt you belonged in school [reversed] .03 .97 5. Were suspended, expelled, had detention .44 .08 6. Had little or no interest in school .86 .34 7. Felt safe at school [reversed] .24 .49 Eigenvalue = 1.98 1.73 Variance = 28.3 24.8 Promax factor correlations: 1 with 2 = .461 Chi-square = 4.02, df = 6, p = 0.67; RMSEA = 0.000

PAGE 267

256 Appendix B: Zero-Order Correlati on Matrix for Final Measures Variable 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 1. Family Disruption T1 2. Family Abuse T1 .566** 3. Poor Peer Relations T1 .328** .131 4. Low Parent Attachment T1 .860** .582** .315** 5. Low School Attachment T1 -.003 -.036 -.192* .014 6. Low School Commitment T1 .060 .104 -.111 .077 .729** 7. Delinquent Peers T1 .514** .319** .369** .534** -.133 -.056 8. Family Disruption T2 .487** .326** 222** .371** .031 .124 .317** 9. Family Abuse T2 .038 .186* -.128 .047 .042 .120 .073 .171* 10. Poor Peer Relations T2 .275** .159 .455 ** .313** -.156 -.132 .333** .178* -.024 11. Low Parent Attachment T2 .434** .346** .251** .408** -.087 .058 .374** .536** .309** .421** 12. Low School Attachment T2 -.018 -.059 -.050 -.059 .161 .195* -.107 .046 .066 -.068 .053 13. Low School Commitment T2 -.042 -.040 -.113 -.064 .148 .219* -.120 .004 .100 -.111 .084 14. Delinquent Peers T2 .327** .187* .334** .382** -.205* -.120 .550** .277** .076 .592** .460** Note p < .05. ** p < .01. (Continued on the next page)

PAGE 268

257 Appendix B: (Continued) Variable 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 15. Delinquency (log) T1 .298** .180* .264** .379 ** -.074 .034 .433** .340** .165 .151 .409** 16. Delinquency (log) T2 .125 .160 .135 .147 -.023 .094 .347** .242** .349** .204* .414** 17. Drug Problems T1 .318** .188* .067 .364 ** -.100 -.025 .447** .157 .133 .183* .309** 18. Drug Problems T2 .283** .224** .076 .312** -.011 .013 .394** .303** .190* .254** .414** 19. Externalizing T1 .534** .326** .294** .500** -.032 .043 .534** .391** .099 .304** .496** 20. Internalizing T1 .398** .313** .463** .441** -.184* -.130 .292** .284** -.063 .367** .381** 21. Externalizing T2 .257** .201* .289** .276** -.109 -.033 .374** .272** .190* .378** .527** 22. Internalizing T2 .222** .112 .335** .238** -.134 -.053 .275** .275** .127 .518** .559** 23. APSD Impulsivity T1 .345** .192* .206* .370** .038 .107 .466** .187* .143 .160 .401** 24. YPI ImpulsivityIrresponsibility T1 .470** .291** .242** .472** .012 .140 .544** .300** .082 .300** .440** 25. YPI Impulsivity T1 .378** .204* .204* .368** .027 .117 .412** .234** .017 .230** .348** 26. YPI Irresponsibility T1 .323** .209* .272** .307** .007 .115 .457** .262** .093 .388** .379** 27. YPI Thrill-Seeking T1 .458** .306** .127 .488 ** -.003 .115 .478** .248** .093 .133 .361** Note p < .05. ** p < .01. (Continued on the next page)

PAGE 269

258 Appendix B: (Continued) Variable 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 1. Family Disruption T1 2. Family Abuse T1 3. Poor Peer Relations T1 4. Low Parent Attachment T1 5. Low School Attachment T1 6. Low School Commitment T1 7. Delinquent Peers T1 8. Family Disruption T2 9. Family Abuse T2 10. Poor Peer Relations T2 11. Low Parent Attachment T2 12. Low School Attachment T2 13. Low School Commitment T2 .919** 14. Delinquent Peers T2 -.122 -.117 Note p < .05. ** p < .01. (Continued on the next page)

PAGE 270

259 Appendix B: (Continued) Variable 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 15. Delinquency (log) T1 .046 .010 .291** 16. Delinquency (log) T2 -.111 -.144 .405** .403** 17. Drug Problems T1 -.081 -.085 .338** .271** .226** 18. Drug Problems T2 .037 -.011 .383** .296** .411** .486** 19. Externalizing T1 -.007 -.075 .428** .456** .332** .325** .294** 20. Internalizing T1 -.060 -.074 .400** .205* .009 .273** .147 .453** 21. Externalizing T2 -.162 -.209* .494** .419** .621** .325** .481** .545** .256** 22. Internalizing T2 -.099 -.124 .476** .134 .296** .288** .356** .301** .457** .550** 23. APSD Impulsivity T1 -.101 -.097 .383** .422** .312** .367** .250** .546** .308** .445** .220** 24. YPI ImpulsivityIrresponsibility T1 .023 -.074 .404** .483** .404** .397** .338** .646** .323** .498** .307** 25. YPI Impulsivity T1 .046 -.051 .352** .388** .314** .260** .229** .540** .258** .396** .223** 26. YPI Irresponsibility T1 -.006 -.097 .373** .362** .3 32** .431** .372** .476** .273** .425** .357** 27. YPI Thrill-Seeking T1 .016 -.037 .278** .442** .352** .290** .241** .580** .270** .412** .186* Note p < .05. ** p < .01. (Continued on the next page)

PAGE 271

260 Appendix B: (Continued) Variable 23. 24. 25. 26. 27. 1. Family Disruption T1 2. Family Abuse T1 3. Poor Peer Relations T1 4. Low Parent Attachment T1 5. Low School Attachment T1 6. Low School Commitment T1 7. Delinquent Peers T1 8. Family Disruption T2 9. Family Abuse T2 10. Poor Peer Relations T2 11. Low Parent Attachment T2 12. Low School Attachment T2 13. Low School Commitment T2 14. Delinquent Peers T2 Note p < .05. ** p < .01. (Continued on the next page)

PAGE 272

261 Appendix B: (Continued) Variable 23. 24. 25. 26. 27. 15. Delinquency (log) T1 16. Delinquency (log) T2 17. Drug Problems T1 18. Drug Problems T2 19. Externalizing T1 20. Internalizing T1 21. Externalizing T2 22. Internalizing T2 23. APSD Impulsivity T1 24. YPI ImpulsivityIrresponsibility T1 .668** 25. YPI Impulsivity T1 .598** .849** 26. YPI Irresponsibility T1 .413** .784** .494** 27. YPI Thrill-Seeking T1 .634** .840** .599** .466** Note p < .05. ** p < .01.

PAGE 273

262 About the Author Jennifer Wareham received both a Bachel or of Arts Degree and a Master’s Degree in Criminology from the University of South Florida in 1998 and 2001, respectively. She entered the Doctoral Pr ogram in Criminology at the University of South Florida in 2001, and earned a Graduate Certificate in Ge ographic Information Systems in 2003. Since 1998, she has worked as a Resear ch Assistant on several departmental grants and worked as Statistician on two fe derally funded grants. She has coauthored articles in the Journal of Offender Rehabilitation and the Journal of Crime and Justice and was lead author of a publication in Western Criminology Review She is also coauthor of manuscripts currently in press in the Journal of Child and Adolescent Substance Abuse and Criminal Justice and Behavior In addition, she is the Co-Principal Investigator of a study examining violen t behavior among men attending domestic violence batterer court-mandated pr ograms in Hillsborough County, Florida.