Do Nutritional Factors Influence Externalizing Behavior During Early Childhood? : a Genetically Informed Analysis Dylan B

Florida State University Libraries

Electronic Theses, Treatises and Dissertations The Graduate School

2015 Do Nutritional Factors Influence Externalizing Behavior during Early Childhood? : A Genetically Informed Analysis Dylan B. (Dylan Baker) Jackson

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected] FLORIDA STATE UNIVERSITY

COLELGE OF CRIMINOLOGY AND CRIMINAL JUSTICE

DO NUTRITIONAL FACTORS INFLUENCE EXTERNALIZING BEHAVIOR

DURING EARLY CHILDHOOD? A GENETICALLY INFORMED ANALYSIS

DYLAN B. JACKSON

A Dissertation submitted to the College of Criminology and Criminal Justice in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Degree Awarded: Summer Semester, 2015

Dylan B. Jackson defended this dissertation on May 22, 2015.

The members of the supervisory committee were:

Kevin M. Beaver

Professor Directing Dissertation

Amy M. Burdette

University Representative

William D. Bales

Committee Member

Brian J. Stults

Committee Member

The Graduate School has verified and approved the above-named committee members, and

certifies that the dissertation has been approved in accordance with university requirements.

For Emily and Roman

iii ACKNOWLEDGEMENTS

I would first like to acknowledge my mentor and supervisor, Dr. Kevin M. Beaver. I am incredibly grateful for his encouragement and support while working on this dissertation. During my time here at Florida State, Dr. Beaver has opened my mind to the complexity of the origins of human behavior, particularly antisocial behavior. He has also instilled within me a greater confidence in my ability to conduct research and to think outside the box. His insights and guidance have been invaluable during my time as a graduate student. I am inspired by his intellect and academic rigor and I look forward to our future research collaborations.

I would also like to thank my dissertation committee for their support and service. Dr.

William Bales, Dr. Amy Burdette , and Dr. Brian Stults have each provided valuable feedback on previous drafts of this dissertation, for which I am incredibly grateful. I am very humbled by their willingness to dedicate their time and energy to this project and am thankful for their kindness and reassurance throughout this process.

I would also like to acknowledge the College of Criminology and Criminal Justice for affording me the opportunity to receive an exemplary graduate education and for providing me with a research assistantship during my time at Florida State. Additionally, I would like to thank

The Graduate School at Florida State University for offering me the university fellowship during my first 3 years as a graduate student. Both the research assistantship and the university fellowship facilitated the successful completion of my graduate education, and I am incredibly grateful to the college and The Graduate School for their generosity.

Finally, I would like to thank my family and friends for their love and encouragement during my years in graduate school. Those closest to me never doubted me when I decided to embark on this journey, even when I doubted myself. I would like to express my deepest love

iv and thanks to my parents, Lee and Nancy Jackson. They have dedicated countless hours of their life to my success and happiness, and I hope they know that all of their efforts have not gone unnoticed. Most importantly, they have taught me how to be a man of kindness and good moral character, for which I will forever be grateful. I am also grateful to my dear son Roman, for greeting me with open arms and a big smile every night during the past few years. He is the source of such immense joy and purpose in my life, and I know that my time in graduate school would have been far more arduous had he not been there to put a smile on my face and to warm my heart.

Most of all, I would like to thank my wife Emily for her love and companionship as well as her unwavering commitment to my happiness, well-being, and success. I will never be able to repay her kindness and patience with me during this stage of our life together. She has endured the thick and thin of my graduate education over the past 6 years, and I am under no illusion that any of this would have been possible without her perpetual support. She is the bedrock of my happiness, my hope for a wonderful future, and my reason for living.

v TABLE OF CONTENTS

LIST OF TABLES ...... viii LIST OF FIGURES ...... xi ABSTRACT ...... xii

CHAPTER 1 STATEMENT OF THE PROBLEM ...... 1 1.1 Research Questions ...... 5 1.2 Outline...... 8

CHAPTER 2 THE EARLY CHILDHOOD ORIGINS OF ANTISOCIAL BEHAVIOR ...... 9 2.1 Moffitt’s (1993) Developmental Taxonomy ...... 9 2.2 Childhood Externalizing Behavior as a Predictor of Offending ...... 12 2.3 The Etiology of Childhood Externalizing Behavior ...... 15

CHAPTER 3 NUTRITION: A VOID IN THE LITERATURE ...... 21 3.1 Infant Nutrition ...... 22 3.2 Early Childhood Nutrition ...... 25 3.3 The Association between Nutritional Factors and Attachment Security ...... 27 3.4 Key Limitations in the Nutrition Literature ...... 31 3.5 Contribution of the Current Study ...... 42

CHAPTER 4 NUTRITION AND EXTERNALIZING BEHAVIOR: A MULTIDISCIPLINARY REVIEW ...... 45 4.1 Breastfeeding and Externalizing Behavior ...... 46 4.2 Breastfeeding: Optimal Infant Nutrition or Facilitator of Attachment Security? ...... 51 4.3 Attachment Security and Externalizing Behavior ...... 55 4.4 Early Childhood Nutrition and Externalizing Behavior ...... 58 4.5 Is Early Childhood Nutrition Associated with Attachment Security? ...... 64 4.6 The Role of Genetic Factors in the Development of Externalizing Behavior ...... 66 4.7 The Role of Perinatal Factors in the Development of Externalizing Behavior ...... 68

vi 4.8 A Biosocial Process: Differential Vulnerability to Poor Nutrition ...... 70 4.9 Nutritional Interplay and the Formation of Externalizing Behavior ...... 76

CHAPTER 5 METHODS ...... 79 5.1 Data ...... 79 5.2 Measures: Breastfeeding Analyses ...... 82 5.3 Measures: Low Diet Quality Analyses ...... 93 5.4 Plan of Analysis ...... 100

CHAPTER 6 RESULTS ...... 109 6.1 Results: Breastfeeding Analyses ...... 109 6.2 Results: Low Diet Quality Analyses ...... 134 6.3 Results: Nutritional Interplay Analyses ...... 154

CHAPTER 7 DISCUSSION ...... 218 7.1 Summary of Results ...... 218 7.2 Policy Implications ...... 233 7.3 Limitations of the Current Study ...... 240 7.4 Future Directions for Theory and Research ...... 244

APPENDICIES ...... 247 A. ITEMS FOR NUTRITION MEASURES ...... 247 B. ITEMS FOR THE LOW ATTACHMENT SECURITY MEASURE ...... 249 C. ITEMS FOR THE EXTERNALIZING BEHAVIOR MEASURE...... 256 D. IRB APPROVAL FORMS ...... 258 REFERENCES ...... 262 BIOGRAPHICAL SKETCH ...... 297

vii LIST OF TABLES

Table 6.1: Descriptive Statistics for the Same-Sex Twin Sample (Breastfeeding Analyses) ...... 166

Table 6.2: Bivariate Correlations (Breastfeeding Analyses) ...... 167

Table 6.3: Child and Maternal/Household Profiles across Breastfeeding Threshold (6 mos.) ....168

Table 6.4: Child and Maternal/Household Profiles by Exclusivity of Breastfeeding...... 169

Table 6.5: Descriptive Statistics of the Variables used in the Defries-Fulker Analyses (Breastfeeding Analyses) ...... 170

Table 6.6: DF analysis of the Shared Environment, Heritability, and Breastfeeding Practices as Predictors of Externalizing Behavior during Kindergarten ...... 171

Table 6.7: DF analysis of the Shared Environment, Heritability, and Breastfeeding Practices as Predictors of Low Attachment Security during Toddlerhood...... 173

Table 6.8: DF analysis of the Shared Environment, Heritability, and Low Attachment Security as Predictors of Externalizing Behavior during Kindergarten ...... 174

Table 6.9: Does Low Attachment Security Explain the Link between Breastfeeding Practices and Externalizing Behavior?...... 175

Table 6.10: Does Birth Weight Confound the Relationship between Breastfeeding Practices and Externalizing Behavior?...... 176

Table 6.11: Is the Relationship between Breastfeeding Practices and Externalizing Behavior in Offspring Moderated by Genetic Risk? ...... 177

Table 6.12: Is the Relationship between Breastfeeding Practices and Externalizing Behavior in Offspring Moderated by Low Attachment Security? ...... 181

Table 6.13: Is the Relationship between Breastfeeding Practices and Externalizing Behavior in Offspring Moderated by Low Birth Weight?...... 182

Table 6.14: Descriptive Statistics for the Same-Sex Twin Sample (Low Diet Quality Analyses)...... 184

Table 6.15: Bivariate Correlations (Low Diet Quality Analyses) ...... 185

Table 6.16: Child and Maternal/Household Profiles across the 90th Percentile of Low Diet Quality...... 187

viii Table 6.17: Child and Maternal/Household Profiles across the 90th Percentile of Low Vegetable Consumption ...... 188

Table 6.18: Child and Maternal/Household Profiles across the 90th Percentile of Low Fruit Consumption ...... 189

Table 6.19: Child and Maternal/Household Profiles across the 90th Percentile of High Fast Food Consumption ...... 190

Table 6.20: Child and Maternal/Household Profiles across the 90th Percentile of High Sweets Consumption ...... 191

Table 6.21: Child and Maternal/Household Profiles across the 90th Percentile of High Salty Snack Consumption ...... 192

Table 6.22: Child and Maternal/Household Profiles across the 90th Percentile of High Soda Consumption ...... 193

Table 6.23: Descriptive Statistics of the Variables used in the Defries-Fulker Analyses (Low Diet Quality Analyses)...... 194

Table 6.24: DF analysis of the Shared Environment, Heritability, and Preschool Dietary Factors as Predictors of Externalizing Behavior ...... 195

Table 6.25: DF analysis of the Shared Environment, Heritability, and Low Attachment Security as Predictors of Dietary Factors ...... 196

Table 6.26: Do the Effects of Dietary Factors on Externalizing Behavior Persist Independent of Low Attachment Security? ...... 197

Table 6.27: Do the Effects of Dietary Factors on Externalizing Behavior Persist Independent of Stability in Externalizing Behavior? ...... 198

Table 6.28: Is the Relationship between Dietary Factors and Externalizing Behavior Moderated by Genetic Risk? ...... 199

Table 6.29: Is the Relationship between Dietary Factors and Externalizing Behavior Moderated by Low Attachment Security? ...... 202

Table 6.30: Descriptive Statistics of the Nutritional Factors across Life Stages ...... 204

Table 6.31: Bivariate Correlations of Nutritional Factors across Life Stages ...... 205

Table 6.32: Breastfeeding Profiles across the 90th Percentile of Preschool Dietary Factors ...... 206

ix Table 6.33: DF analysis of the Shared Environment, Heritability, and Short Duration of Breastfeeding as Predictors of Dietary Factors ...... 208

Table 6.34: DF analysis of the Shared Environment, Heritability, and Breastfed Less than 6 Months as Predictors of Dietary Factors ...... 209

Table 6.35: DF analysis of the Shared Environment, Heritability, and Not Exclusively Breastfed as Predictors of Dietary Factors ...... 210

Table 6.36: Is the Breastfeeding Threshold Effect on Externalizing Behavior Explained by Preschool Dietary Factors? ...... 211

Table 6.37: Do Breastfeeding Practices and Preschool Dietary Factors Interact to Predict Externalizing Behavior in Offspring? Summary of Findings ...... 213

Table 6.38: Is the Relationship between Breastfeeding Practices and Externalizing Behavior in Offspring Moderated by Low Diet Quality? ...... 217

x LIST OF FIGURES

Figure 6.1: An Illustration of the Breastfeeding Threshold Effect within Hypothetical Twin Pairs...... 172

Figure 6.2: The Predicted Externalizing Behavior Score (W4/5) By Level of Genetic Risk and Short Duration of Breastfeeding When Covariates Are at Their Mean ...... 178

Figure 6.3: The Predicted Externalizing Behavior Score (W4/5) By Breastfeeding Threshold (6 Months) and Level of Genetic Risk When Covariates Are at Their Mean ...... 179

Figure 6.4: The Predicted Externalizing Behavior Score (W4/5) By Exclusivity of Breastfeeding (>= 6 Months) and Level of Genetic Risk When Covariates Are at Their Mean ...... 180

Figure 6.5: The Predicted Externalizing Behavior Score (W4/5) By Breastfeeding Threshold (6 Months) and Birth Weight Status When Covariates Are at Their Mean ...... 183

Figure 6.6: The Predicted Externalizing Behavior Score (W4/5) By Level of Genetic Risk and Fast Food Consumption When Covariates Are at Their Mean ...... 201

Figure 6.7: An Illustration of the Direct and Indirect Effects of Breastfeeding Practices on Externalizing Behavior ...... 212

Figure 6.8: The Predicted Externalizing Behavior Score (W4/5) By Low Vegetable Consumption and Breastfeeding Threshold (6 Months) When Covariates Are at Their Mean ...... 214

Figure 6.9: The Predicted Externalizing Behavior Score (W4/5) By High Salty Snack Consumption and Breastfeeding Threshold (6 Months) When Covariates Are at Their Mean ...215

Figure 6.10: The Predicted Externalizing Behavior Score (W4/5) By High Sweets Consumption and Exclusive Breastfeeding When Covariates Are at Their Mean ...... 216

Figure B.1: Illustration of the Profile of the Secure Attachment Prototype Across Hotspots .....254

Figure B.2: Illustration of the Profile of the Insecure Attachment Prototype Across Hotspots ..255

xi ABSTRACT

Moffitt’s (1993) taxonomy of adolescence-limited and life-course-persistent offenders suggests, among other things, that an early onset of antisocial behavior a) increases the likelihood of a life- course persistent offending trajectory and b) is the product of neuropsychological deficits and familial risk factors working in concert. Over two decades of research has yielded a substantial amount of support for Moffitt’s claims. Nevertheless, research has yet to significantly expand the repertoire of biosocial processes that might influence the onset of antisocial behavior during childhood. For instance, only a handful of criminologists have considered the role that infant and early childhood nutrition might play in the development of antisocial behavior. Moreover, studies employing genetically sensitive methods to examine the link between nutrition and child antisocial behavior are virtually nonexistent. Scholars have also neglected to consider the socioemotional mediators of the nutrition-externalizing relationship, as well as whether infants and children are differentially sensitive to nutritional intake depending on their level of genetic risk, perinatal risk, and security of attachment. Finally, researchers have yet to explore the ways in which nutritional factors across infancy and early childhood work together to influence externalizing behavior. This dissertation seeks to address these gaps in the literature by employing a large, nationally representative sample of twin pairs. The findings suggest that, even after accounting for the influence of genes and the shared environment, nutritional factors during infancy and early childhood have both direct and indirect effects on externalizing behavior during kindergarten. Furthermore, some significant interactions between genetic and nutritional factors, as well as nutritional factors across life stages, were detected. The limitations of the study are noted and recommendations for policy, theory, and future research are discussed.

xii CHAPTER 1

STATEMENT OF THE PROBLEM

There is no doubt that environments [such as] family, peers, [and the] neighborhood play an important role in the development of antisocial behavior. However, by neglecting to prospectively study the development of antisocial behavior from early childhood onwards and to take account of biological mechanisms, criminologists will remain blind to the developmental origins of antisocial behavior.

-Richard E. Tremblay (2015, p. 40)

Since the inception of the field, the overwhelming majority of criminologists, in their efforts to understand the etiology of crime, have limited their theorizing and empirical explorations to behaviors and contextual factors that emerge following the onset of adolescence.

Such a focus is understandable, considering the increased prevalence of offending during adolescence and the persistence of serious crime into adulthood among a smaller subset of individuals. Over the past couple of decades, however, a growing number of scholars have attempted to widen the scope of criminological theory and research to include examinations of environments, behaviors, and traits that are present during earlier life stages (Beaver & Wright

2005; Farrington, 2003; Jackson & Beaver, 2013; Moffitt & Caspi, 2001; Piquero, 2001; Tibbetts

& Piquero, 1999). This body of work is owed in large part to the theorizing of Moffitt (1993), a prominent life-course scholar who argued that neuropsychological deficits, coupled with criminogenic environments within the family (e.g., adverse parenting), underpin an early onset of offending, and that an early onset of offending increases the risk of chronic, serious offending into adulthood. Her contentions inspired a number of empirical inquiries about the relevance of

1 childhood processes for later criminal behavior (Caspi, Henry, McGee, Moffitt, & Silva, 1995;

Kratzer & Hodgins, 1999; Moffitt & Caspi, 2001; Piquero, 2001; Tibbetts & Piquero, 1999).

This body of literature has revealed that a) childhood onset of antisocial behavior is one of the best predictors of a life-course-persistent offending trajectory (Broidy et al., 2003; Nagin &

Tremblay, 1999; Tremblay et al., 2004) and b) early-onset, life-course-persistent antisocial behavior is influenced by a combination of both neuropsychological (Moffitt & Caspi, 2001;

Piquero, 2001; Raine et al., 2005) and familial factors (Tremblay et al., 2004; Turner, Hartman,

& Bishop, 2007).

Despite these rather robust findings, scholars have been slow to identify predictors of an early onset of offending beyond what Moffitt’s theory specifically postulates. In short, researchers have not sufficiently expanded the developmental predictors of childhood antisocial behavior to include other potentially important biosocial processes, such as poor nutritional intake during the first few years of life and genetic resilience or sensitivity to such intake. This gap in the literature is especially noteworthy in light of recent research suggesting that life- course-persistent offending is at least partly genetic in origin (Barnes, 2013; Barnes, Beaver, &

Boutwell, 2011) and that genes can make individuals more or less sensitive to risk and protective factors in the environment (Belsky & Beaver, 2011). Additionally, various studies in the fields of medicine, psychiatry, public health, epidemiology, and child development have suggested that nutritional factors at the earliest stages of life are related to subsequent development, both cognitive (Belfort et al., 2013; Gómez-Pinilla, 2008; Julvez et al., 2014; Molteni, Barnard, Ying,

Roberts, & Gómez-Pinilla, 2002) and behavioral (Galler et al., 2011; Heikkilä, Sacker, Kelly,

Renfrew, & Quigley, 2011; Julvez et al., 2007; Woo et al., 2014). Despite these findings, criminologists have not adequately incorporated this body of work into criminological theory or

2 research and, as a result, have overlooked its potential to inform preventive strategies. In short,

Moffitt’s theory has not yet been amended to acknowledge the findings of interdisciplinary research regarding the role of nutritional factors in child antisocial behavior. More broadly, greater attention to multidisciplinary research, along with novel explorations of potentially important biosocial interactions, is vital in order for criminologists to more fully capture the various environmental, biological, and physiological factors that can play a role in the development of antisocial behavior during childhood (Tremblay, 2015).

To be precise, there are four overarching limitations in the current body of criminological research that examines the role of nutritional factors in the development of childhood antisocial behavior. First, apart from the literature supporting Moffitt’s claims, our knowledge regarding the precursors of childhood-onset antisocial behavior is still quite limited, especially in relation to our knowledge concerning other topics of criminological inquiry (e.g., the age-crime curve, crime correlates, etc.). This gap in our understanding is at least partly the result of the unwillingness and/or inability of criminologists to incorporate the hypotheses and the findings from multidisciplinary research on childhood-onset antisocial behavior, including research that has linked nutritional factors to child antisocial behavior, into mainstream criminology. Second, research to date that has explicitly examined nutritional precursors to childhood antisocial behavior a) has almost exclusively been conducted outside the field of criminology and b) typically does not account for the confounding and/or moderating role of genetic/biological risk factors in the link between nutrition and childhood antisocial behavior (Galler et al., 2011;

Heikkilä et al., 2011; Julvez et al., 2007; Oh, Ahn, Chang, Kang, & Oh, 2013; Woo et al., 2014).

In short, most scholars have overlooked the utility of a genetically informative approach when studying the link between nutrition and child antisocial behavior. This oversight is an important

3 one, because when implemented, such an approach can “play a unique role in bridging gaps between biological and social science research” by a) allowing researchers to determine the extent to which a childhood onset of antisocial behavior is biosocial in nature and b) enhancing researchers’ ability to distinguish between environmental correlates and causes of childhood antisocial behavior (D’Onofrio, Lahey, Turkheimer, & Lichtenstein, 2013, p. S46; see also

Jaffee, Strait, & Odgers, 2012). The relevance of this gap in the literature cannot be overstated, as studies exploring the interplay between nutritional, developmental, and genetic factors in the prediction of childhood antisocial behavior are virtually absent from the literature.

Third, scholars have not adequately explored the developmental processes and events that link nutritional risk factors during infancy and early childhood to an early onset of antisocial behavior. The few scholars that have examined mediating mechanisms tend to focus on cognitive and/or neuropsychological processes (see Liu, Raine, Venables, & Mednick, 2004, for an example). Nutritional patterns during infancy and early childhood, however, have been found to both influence and reflect the socioemotional connection, or the level of attachment, of the child to their caregiver (Bost et al., 2014; Britton, Britton, Gronwaldt, 2006; Tharner et al.,

2012). Thus, it is entirely possible that the security of the infant-caregiver attachment may a) mediate the relationship between infant nutrition (e.g., breastfeeding) and childhood antisocial behavior and/or b) confound the relationship between early childhood nutrition and childhood antisocial behavior. Research to date, however, has yet to explore these possibilities. More broadly, the specific biosocial processes linking nutritional intake to childhood antisocial behavior are not fully understood beyond the effect of acute malnutrition on IQ (Liu et al., 2004).

Finally, nutrition scholars have traditionally examined infant nutrition and early childhood nutrition in isolation (Julvez et al., 2007; Wiles, Northstone, Emmett, & Lewis, 2007). More

4 recently, however, a handful of researchers have begun to explore the possibility that the quality of nutrition across these two life stages may be linked (Perrine, Galuska, Thompson, & Scanlon,

2014; Scott, Chih, & Oddy, 2012). Still, this emerging body of research does not consider the ways in which infant and child nutrition might work together to increase the risk of childhood antisocial behavior. For example, it is possible that infant nutrition has no effect on externalizing behavior independent of subsequent nutritional intake during childhood, or that the influence of nutritional factors at one life stage may be contingent on nutritional intake at another life stage.

A deeper understanding of the role of nutrition across both life stages in the development of childhood antisocial behavior can provide insight into which nutritional factors are more (or less) important to target, with the ultimate goal of preventing and/or minimizing child antisocial behavior.

1.1 Research Questions

In an effort to a) expand the repertoire of developmental processes that might underpin childhood antisocial behavior and b) uncover some potentially important oversights in Moffitt’s

(1993) theory, this dissertation will explore the ways in which nutritional factors across infancy and early childhood (e.g., breastfeeding, diet quality) can influence the onset of antisocial behavior during childhood. Put simply, the current dissertation takes a developmental approach to the origins of antisocial behavior. Such an approach focuses on “identifying risk factors for children” and seeks “to improve environmental conditions in order to facilitate [the] healthy development of the child.” (Rocque, Welsh, & Raine, 2012, p. 310). The current dissertation also acknowledges that genetic factors may make children differentially sensitive to early childhood environments such as nutrition. As a result, genetically informative techniques are

5 utilized in order to permit empirical testing of nutrition-related biosocial processes. Ultimately, I aim to directly address several important theoretical and methodological limitations in the criminological literature, namely 1) the lack of criminological research that builds upon the precursors to childhood antisocial behavior designated by Moffitt (1993) and situates the findings of interdisciplinary research into mainstream criminological inquiry, 2) the absence of studies that consider the role of genetic/biological risk factors in the relationship between nutritional factors and childhood antisocial behavior, 3) the paucity of research exploring the mechanisms, particularly the socioemotional mechanisms, that might link nutritional factors during infancy and early childhood to childhood antisocial behavior, and 4) the lack of attention to the interplay of nutritional factors across infancy and early childhood in the prediction of childhood antisocial behavior.

The research questions that will be examined in this dissertation will be divided into three sections. The first set of research questions is limited to questions examining the significance of nutritional factors during infancy, or more precisely, the duration and exclusivity of breastfeeding. The second set of research questions is limited to questions examining the significance of nutritional factors during early childhood, or more precisely, the quality of the diet during preschool across six dimensions. The third and final set of research questions examine the interplay between nutritional factors across infancy and early childhood. The research questions are as follows:

Infant Nutrition:

1. In what ways do mothers and children who breastfeed for shorter durations differ from

mothers and children who breastfeed for longer durations?

6 2. Does a short duration of breastfeeding significantly increase the risk of low attachment

security during toddlerhood and/or externalizing behavior problems during kindergarten,

independent of familial and genetic influences?

3. Is the relationship between short duration of breastfeeding and externalizing behavior

problems during kindergarten explained by low attachment security during toddlerhood

and/or low birth weight?

4. Is the influence of breastfeeding duration on kindergarten externalizing behavior

moderated by genetic risk, low attachment security, and/or low birth weight?

Early Childhood Nutrition:

5. In what ways do children with especially poor dietary habits differ from other children?

6. Does a low quality diet during preschool significantly increase the risk of externalizing

behavior problems during kindergarten, independent of familial and genetic influences?

7. Is the relationship between a low quality diet during preschool and externalizing behavior

during kindergarten robust to indicators of low attachment security during toddlerhood

and externalizing behavior during preschool?

8. Is the influence of a low quality diet on kindergarten externalizing behavior moderated by

genetic risk and/or low attachment security?

Interplay Between Nutritional Factors:

9. Does a short duration of breastfeeding increase the likelihood of a low quality diet?

10. Is the relationship between short duration of breastfeeding and externalizing behavior

problems during kindergarten explained by a low quality diet during preschool?

11. Do a short duration of breastfeeding and a low quality diet during preschool interact to

predict kindergarten externalizing behavior?

7 1.2 Outline

Due to the number and the multidisciplinary nature of the proposed research questions, this dissertation will cover a broad array of literature. As a result, a brief overview of the organization of this dissertation is warranted before proceeding further. Chapter 2 discusses the theories and research that are relevant to childhood antisocial behavior. Chapter 3 provides an introduction to the literature on nutrition during infancy and early childhood, discusses the link between attachment security and these nutritional factors, proposes five key limitations of the nutrition literature, and outlines the contribution of the current study to the literature. Chapter 4 present a thorough, multidisciplinary review of the literature that examines the relationship between nutritional factors and externalizing behavior, links genetic and developmental factors to the formation of externalizing behavior, and highlights the relevance of the interplay between biological and environmental factors in the formation of externalizing behavior. Chapter 5 outlines the methods relevant to the execution of the present study, including the data, measures, and the plan of analysis. Chapter 6 presents the results pertaining to each of the eleven questions posited in the previous section of this introductory chapter. Finally, this dissertation will conclude with a summary of the findings, as well as a discussion of the policy implications of the current study, the limitations of the current study, and possible directions for future theory and research, all of which are presented in chapter 7.

8 CHAPTER 2

THE EARLY CHILDHOOD ORIGINS OF ANTISOCIAL BEHAVIOR

Over the last two decades, a number of criminologists have argued that developmental and environmental processes that precede adolescence can influence the likelihood of criminal offending (Farrington, 2003; Gottfredson & Hirschi, 1990; Moffitt, 1993; Sampson & Laub,

1993). Although many of the details of these theoretical developments differ, they all acknowledge that environmental and/or individual factors during the early stages of the life course can have a tremendous impact on whether an individual will subsequently engage in various forms of delinquency. For instance, Gottfredson and Hirschi’s (1990) general theory of crime suggests that both parental monitoring and discipline during the early years of life play a pivotal role in the development of self-control, and that the failure to develop adequate self- control during childhood increases the risk of subsequent delinquency. Conversely, Sampson and Laub’s (1993) age-graded theory of informal social control argues that social bonds across the life course, whether with family, friends, colleagues, or institutions, can both a) diminish the likelihood of offending and b) facilitate desistance among offenders. Thus, while Sampson and

Laub (1993) do not subscribe to the notion of a latent trait (e.g., low self-control) that underpins all delinquent activity, they do acknowledge that dynamic environmental factors at various life stages can impact delinquent trajectories.

2.1 Moffitt’s (1993) Developmental Taxonomy

One of the most prolific of the life-course theories is Moffitt’s (1993) developmental taxonomy of adolescence-limited and life-course-persistent offending. In her theory, Moffitt

(1993) argues that there are two distinct groups of offenders, in addition to a group of non-

9 offenders (i.e., abstainers). The two groups of offenders are referred to as adolescence-limited

(AL) and life-course persistent (LCP) offenders. Moffitt’s unique theoretical contribution centers around her contention that the etiologies, trajectories, and patterns of offending are markedly different between AL offenders and LCP offenders.

According to the theory, AL offenders engage in more minor, transient forms of delinquency that are undetectable prior to adolescence. The deviant acts of AL offenders are thought to arise from a) efforts to mimic LCP offenders and b) the “maturity gap”, which is defined as “a time warp between biological age and social age” (p. 687). To be precise, AL offending is conceptualized as a non-pathological means of asserting individual autonomy in accordance with biological development and in the face of social restrictions imposed by parents and society at large. Consequently, AL offending is expected to desist upon becoming an adult due to enhanced social autonomy and less frequent contact with (and less desire to imitate) LCP offenders. In short, the primary motivations that underpin AL offending (i.e., association with

LCP offenders, the maturity gap) are expected to shift significantly upon entering adulthood, which, according to the theory, would result in the desistance of AL offenders by early adulthood.

The second group of offenders (i.e., LCP offenders) exhibits an entirely different offending trajectory, one that has its origins in the earliest stages of the life course. According to the theory, LCP offenders begin to manifest antisocial behaviors during early childhood. Moffitt

(1993) also posits that the antisocial behaviors that characterize LCP offenders a) are caused by an interactional continuity of individual and familial risk factors that are present during the earliest stages of life (e.g., neuropsychological deficits, difficult temperament, and family risk) and b) are likely to remain stable through adolescence and adulthood. Thus, a biosocial process,

10 in which personal neuropsychological disadvantage and adverse home environments are continually reinforcing each other, is the proposed cause of the LCP offending trajectory. This process is thought to ultimately result in a pathological personality, which is expected to culminate in a pattern of chronic, and even violent, offending.

Moffitt’s taxonomy has garnered considerable empirical attention over the past couple of decades. Overall, empirical research examining the tenets of her theory has been largely supportive (for thorough reviews, see Jennings & Reingle, 2012; Moffitt, 2006). Although scholars have not consistently found evidence of exactly two distinct offending groups and one abstaining group (Blokland, Nagin, & Nieuwbeerta, 2005; Brame, Bushway, & Paternoster,

2003; Nagin & Land, 1993; Nagin & Tremblay, 1999; Odgers et al., 2008), the offending patterns of the groups detected are largely consistent with Moffitt’s taxonomy (Jennings &

Reingle, 2012; Piquero, 2008). The most consistent finding across studies is the discovery a group of high-rate, chronic LCP offenders (Piquero, 2008). Specifically, research has repeatedly yielded evidence of a small, chronic offending group, which typically comprises about 5% of all offenders, yet commits approximately 50% of all offenses (Nagin & Tremblay, 1999; Piquero,

Farrington, Nagin, & Moffitt, 2010; Vaughn et al., 2011; van der Geest, Blokland, & Bijleveld,

2009; Shaw, Lacourse, & Nagin, 2005).

Furthermore, research has indicated that this small group of LCP offenders are significantly more likely to show evidence of under-controlled temperament (Caspi, Moffitt,

Newman, & Silva, 1996; Henry, Caspi, Moffitt, & Silva, 1996), neurological and motor deficits

(Moffitt & Caspi, 2001; Piquero, 2001; Raine et al., 2005), hyperactivity (Nagin & Tremblay,

1999) and physical aggression (Broidy et al., 2003) during toddlerhood and early childhood.

Research has suggested that these childhood traits are frequently accompanied by a significant

11 degree of risk in the family environment (Moffitt & Caspi, 2001), intimating support for

Moffitt’s (1993) hypothesis concerning the origins of LCP offending. Scholars have also found that LCP offenders have diminished well-being across various life dimensions, including employment, mental health, and family life (Moffitt, Caspi, Herrington, & Milne, 2002). Thus, the evidence for a small group of chronic, LCP offenders, who experience severe life challenges and engage in serious crime through adulthood, is quite robust (Moffitt, 2006). Additionally, there is ample evidence to suggest that these offenders experience significant neuropsychological, temperamental, and familial disadvantage during the early stages of the life course (Broidy et al., 2003; Caspi et al., 1996; Henry et al., 1996; Moffitt & Caspi, 2001;

Piquero, 2001; Raine et al., 2005).

2.2 Childhood Externalizing Behavior as a Predictor of Offending

Importantly, studies testing the validity of Moffitt’s (1993) taxonomy have been accompanied by a wealth of research examining the extent to which an early onset of antisocial behavior predicts stability and severity of future antisocial behavior (Moffitt, Caspi, Dickson,

Silva, & Stanton, 1996; Tremblay, Pihl, Vitaro, & Dobkin, 1994). Life-course criminologists acknowledge the heterotypic continuity of antisocial behavior, which implies that, to measure the extent of stability in antisocial behavior, age-sensitive measures are often required (Wright,

Tibbetts, & Daigle, 2008). For example, although frequent temper tantrums and biting might be appropriate indicators of antisocial behavior during preschool, such items become inappropriate and/or irrelevant measures of antisocial behavior during adolescence. Thus, even if the underlying trait remains stable, the behavioral manifestations that are examined to detect the presence of that trait may not remain stable.

12 When age-appropriate indicators of antisocial behaviors are employed, researchers often find evidence that antisocial children are at risk of becoming antisocial adolescents and adults

(Broidy et al., 2003; Campbell, Shaw, & Gilliom, 2000; Fergusson, Boden, & Horwood, 2014;

Fergusson & Horwood, 1995; Green, Gesten, Greenwald, & Salcedo, 2009; Nagin & Tremblay,

1999; Nagin & Tremblay, 2001; Stattin & Magnusson, 1989). More specifically, elevated levels of physical aggression and associated externalizing behaviors during early childhood seem to greatly increase the odds of criminal activity during adolescence and adulthood (Broidy et al.,

2003; Kokko, Tremblay, Lacourse, Nagin, & Vitaro, 2006; Nagin and Tremblay, 1999;

Thompson et al., 2010). To illustrate, a seminal study by Nagin and Tremblay (1999) examined the physical aggression trajectories of males from childhood to adolescence. The authors found that subjects who exhibited high levels of externalizing behavior at age 6 tended to engage in high levels of delinquency at age 15. The results suggest that early oppositional, externalizing and aggressive behaviors are driven by similar developmental processes, and that this cluster of traits is one of the best predictors of adolescent delinquency (Nagin & Tremblay, 1999). A follow-up study by Nagin and Tremblay (2001) found that the odds of belonging to a “high aggression” group during high school were increased by a factor of 3 for boys who frequently engaged in externalizing behaviors (e.g., hyperactivity, defiance, aggression) during kindergarten.

A number of subsequent studies have corroborated these results. For example, Broidy and colleagues (2003) employed cross-national data to examine whether physical aggression during childhood predicts violent and non-violent offending during adolescence. The results of the study revealed that boys who engaged in chronic physical aggression during elementary school were at high risk of both violent and nonviolent offending during adolescence, although

13 the same was not true for girls. Another study by Kokko and colleagues (2006) examined whether aggression trajectories during childhood predicted the likelihood of school dropout and physical violence by age 17 using a sample of 1,025 males. The results indicated that a small proportion of the boys (3.4%) exhibited chronic, stable trajectories of physical aggression, which was significantly predictive of both school dropout and physical violence.

A recent study by Piquero, Carriaga, Diamond, Kazemian, and Farrington (2012) found that teacher-rated aggression during both childhood and adolescence predicted the likelihood of a criminal conviction through mid-adulthood (age 40), implying a strong degree of continuity in antisocial behavior across the life course. These results are corroborated by research that has indicated that both bullying and externalizing behaviors during childhood are significantly predictive of official and self-report measures of property and violent offending during adulthood

(Fergusson et al., 2014; see also Schaeffer, Petras, Ialongo, Poduska, & Kellam, 2003).

Moreover, other studies have demonstrated that oppositional and hyperactive behaviors during childhood are predictive of drug abuse (Pingault et al., 2012; Timmermans, van Lier, & Koot,

2008), gambling (Shenassa, Paradis, Dolan, Wilhelm, & Buka, 2012), and risky sexual behavior

(Timmermans et al., 2008).

Despite the clear relevance of childhood behavioral problems for subsequent offending, scholars have been careful to note that the occurrence of externalizing behaviors and physical aggression during late infancy and early toddlerhood is often normative (Tremblay et al., 2004;

Alink et al., 2006). For example, Alink and colleagues (2006) examined the stability and prevalence of physical aggression in young subjects (N = 2,253) at 12, 24, 36, and 48 months of age. The results suggest that acts of physical aggression can emerge during the first year of life, but that these acts increase dramatically during the second and third years of life, especially

14 among boys. However, physical aggression tends to decline from 36 months onward. The authors explain that, for most infants and toddlers, physical aggression represents a means of a) expressing anger/frustration and b) interacting with and navigating the social world at a relatively rudimentary level. As children develop a theory of mind, empathy, and greater verbal ability, they become increasingly sophisticated at navigating their social world, and physical aggression becomes much less frequent (Alink et al., 2006; see also Thomas & Pope, 2012).

Thus, as a result of normative developmental processes, it is uncommon for a high rate of aggression to persist through preschool and kindergarten (see Tremblay et al., 2004).

Notwithstanding, this pattern of early childhood behavior is among the best predictors of subsequent offending (Nagin & Tremblay, 2001). Consequently, a better understanding of the early life factors that increase the risk of persistent externalizing behavioral problems is pivotal in order to effectively prevent subsequent offending.

2.3 The Etiology of Childhood Externalizing Behavior

In light of the wealth of literature linking persistent externalizing problems during childhood to offending behaviors during adolescence and adulthood, a number of scholars have sought to explore the early life factors that contribute to the development of externalizing behaviors during childhood (Choe, Olson, & Sameroff, 2013; Miner & Clarke-Stewart, 2008;

Plybon & Kliewer, 2001; Ramchandani et al., 2013). Much of the empirical attention in this area of research has been placed on demographic and familial factors, including sex (Deater-Deckard,

Dodge, Bates, & Pettit, 1997; Keiley, Bates, Dodge, & Pettit, 2000; Miner & Clark-Stewart,

2008), low socioeconomic status (Dodge, Pettit, & Bates, 1994; Plybon & Kliewer, 2001), maternal traits (Choe et al., 2013; Connell & Goodman, 2002; Shaw, Kennan, & Vondra, 1994),

15 and inadequate parenting (Buschgens et al., 2009; Castelao & Kroner-Herwig, 2014; Kerr,

Lopez, Olson, & Sameroff, 2004; Pearl, French, Dumas, Moreland, & Prinz, 2014; Ramchandani et al., 2013) as predictors of externalizing behavior. In terms of biological sex, research has overwhelmingly indicated that, while persistent externalizing problems can and do occur among girls, these problems are much more common among boys. Keiley and colleagues (2000), for example, found evidence that males exhibit a greater increase in externalizing behavior trajectories from kindergarten through seventh grade compared to females. Furthermore, a study by Miner and Clark-Stewart (2008) indicated that, independent of various familial factors, male gender is predictive of both parent and teacher-reported externalizing behaviors at age 9. Still, the authors found that male externalizing behaviors were reduced more than female externalizing behaviors in the presence of maternal sensitivity, intimating an interaction between biological sex and the familial environment.

Children of lower socioeconomic status are also more likely to exhibit externalizing behaviors during early childhood (Dodge et al., 1994; Plybon & Kliewer, 2001). To illustrate, a study by Dodge and colleagues (1994) revealed that preschool socioeconomic status predicted teacher-rated and peer-rated aggression during grades 1, 2 and 3. The economic conditions of the surrounding neighborhood might also play a role in the development of externalizing behavior, independent of individual socioeconomic status. Plybon and Kliewer (2001), for instance, found that children living in poor neighborhoods were more likely to exhibit externalizing behavior problems than children living in middle-class or affluent neighborhoods, particularly if crime levels in the neighborhood were moderate to high.

Apart from demographic and neighborhood factors, familial factors seem to play an important role in the formation of externalizing behavioral problems. Research has most

16 frequently explored whether maternal traits and behaviors predict child behavior (Choe et al.,

2013; Connell & Goodman, 2002; Shaw et al., 1994). In a recent study, Choe and colleagues

(2013) found that maternal distress heightens the risk of child externalizing behavioral problems.

More specifically, the results suggest that a high level of maternal distress interferes with the development of child self-regulation, which in turn increases the risk of externalizing behavior.

A meta-analysis garnered similar findings, indicating that children of mothers with mental health problems are more likely to develop higher levels of externalizing behavior (Connell &

Goodman, 2000). Inadequate and/or harsh parenting, in addition to high-risk maternal traits, has also been associated with a greater degree of externalizing behavior in offspring (Buschgens et al., 2009; Castelao & Kroner-Herwig, 2014; Gershoff, 2002; Kerr et al., 2004; Pearl et al., 2014;

Ramchandani et al., 2013). The use of corporal punishment, for example, is predictive of subsequent aggression and behavioral problems in children (Gershoff, 2002; Mulvaney &

Mebert, 2007). Other research has indicated that even disengaged and/or inconsistent parenting, independent of corporal punishment, might also contribute to child behavioral problems

(Castelao & Kroner-Herwig, 2014; Ramchandani et al., 2013). For example, a recent study by

Castelao and Kroner-Herwig (2014) revealed that children are more likely to develop higher levels of externalizing behavior if their parents exhibit a dysfunctional parenting style (e.g., inconsistency, blame, restriction). Importantly, recent work has also highlighted the likelihood of a bidirectional relationship between parenting style and child behavioral problems (Pearl et al., 2014), which implies an overestimation of unidirectional studies of parenting effects.

Although empirical examinations of the influence of demographic and parenting variables on externalizing behavior are most common, a handful of researchers have explored other potential influences, including genetic (Arseneault et al., 2003; Baker, Jacobson, Raine,

17 Lozano, & Bezdjian, 2007; Barnes, Boutwell, Beaver, & Gibson, 2013), prenatal (Ashford, Van

Lier, Timmermans, Cuijpers, & Koot, 2008; Rice et al., 2010; Van den Bergh & Marcoen, 2004), and perinatal (Buschgens et al., 2009; Clark, Woodward, Horwood, & Moor, 2008; Delobel-

Ayoub et al., 2006; Liu, Raine, Wuerker, Venables, & Mednick, 2009) factors. The results suggest that, in addition to demographic and familial factors, genetic, prenatal, and perinatal risk factors may play an equally important role in the development of externalizing behavior. To illustrate, a recent study by Barnes and colleagues (2013) revealed that 78-82% of the variance in externalizing behavioral problems during childhood can be attributed to genetic factors. The results also indicated that additive genetic factors explain a great degree of the covariance between low self-regulation, corporal punishment, and externalizing behavior, intimating that the link between parenting behaviors and offspring externalizing behaviors is likely overestimated in studies that do not model genetic risk appropriately. Research has found that heritability estimates of childhood antisocial behavior are the highest (~ 96%) when multiple reports of various dimensions of problem behavior comprise the measure (see Baker et al., 2007).

Prenatal factors also appear to increase the risk of childhood behavioral problems

(Ashford et al., 2008; Rice et al., 2010; Van den Bergh & Marcoen, 2004). Despite some null findings (see Boutwell & Beaver, 2010; D’Onofrio et al., 2008), both maternal stress (Rice et al.,

2010; Van den Bergh & Marcoen, 2004) and maternal smoking (Ashford et al., 2008) during the prenatal period seem to contribute to the development of behavioral problems in offspring. In particular, one study found that situational anxiety of the mother during pregnancy explained

15% of the variance in externalizing behavior, even after controlling for a host of potential confounders (Van den Bergh & Marcoen, 2004). Some research, moreover, has even suggested

18 that the effect of maternal smoking during pregnancy on externalizing problems may persist into adolescence (Ashford et al., 2008).

A number of studies have also indicated that various perinatal risk factors, including birth complications (Beck & Shaw, 2005; Liu et al., 2009) and low birth weight (Breslau & Chilcoat,

2000; Huhtala et al., 2012; Wadsby, Nelson, Ingemansson, Samuelsson, & Leijon, 2014), may heighten the risk of childhood externalizing problems. For instance, Liu and colleagues (2009) found that infants who experience significant birth complications are more likely to exhibit externalizing behavioral problems at 11 years of age. The results also suggest that low IQ mediates the link between birth complications and externalizing behavior, corroborating the literature that underscores the relevance of neuropsychological functions in the development of externalizing behavior (see Riggs, Blair, & Greenberg, 2004; Schoemaker, Mulder, Deković, &

Matthys, 2013). Additionally, individuals who are born low birth weight are more likely to engage in higher levels of externalizing behavior than normal birth weight individuals (Breslau

& Chilcoat, 2000; Wadsby et al., 2014). For example, a recent study by Wadsby and colleagues

(2014) found that very low birth weight children display more behavioral problems at 7 and 9 years of age than normal birth weight children. Similar associations have been found between birth weight and attention difficulties (Breslau & Chilcoat, 2000) as well as ADHD (Breslau,

Chilcoat, DelDotto, Andreski, & Brown, 1996; Jackson & Beaver, 2015; Mick, Biederman,

Prince, Fischer, & Faraone, 2002).

Importantly, other early childhood traits, such as poor language ability (Menting, Van

Lier, & Koot, 2011), and resistant/challenging temperament (Bates, Pettit, Dodge, & Ridge,

1998; Rubin, Burgess, Dwyer, & Hastings, 2003) also increase the likelihood of childhood behavioral problems, in part by eliciting peer rejection (Beaver, Boutwell, Barnes, Schwartz, &

19 Connolly, 2014; Menting et al., 2011). These results highlight the role of the child as a social actor, and how disadvantageous traits can exacerbate behavioral problems by increasing the risk of negative peer interactions (Beaver et al., 2014) and even academic challenges (Rhoades,

Warren, Domitrovich, & Greenberg, 2011). An insecure attachment to one’s caregiver also appears to increase the likelihood of externalizing behavior (Kochanska & Kim, 2013;

O’Connor, Collins, & Supplee, 2012). Ultimately, the literature as a whole indicates that the origins of childhood behavioral problems are multifaceted and likely involve complex interactions between genetically influenced traits, prenatal and/or perinatal risk factors, and the immediate context of the child (e.g., family, school).

20 CHAPTER 3

NUTRITION: A VOID IN THE LITERATURE

Among the studies that have examined the origins of childhood externalizing behavior, relatively few of them have considered the role of nutritional factors in the development of externalizing behaviors (Heikkilä et al., 2011; Galler et al., 2011; Shelton, Collishaw, Rice,

Harold, & Thapar, 2011; Woo et al., 2014). The general paucity of research in this area is somewhat surprising, considering the substantial body of literature that a) links nutritional factors to brain development (Black, 2008; Gómez-Pinilla, 2008; Grantham-McGregor & Ani,

2001; Innis, 2007) and b) links particular aspects of brain development to externalizing behavior

(Riggs et al., 2004; Schoemaker et al., 2013). To be precise, adequate ingestion of various nutrients, vitamins, and minerals (e.g., folate, zinc, iron, magnesium, polyunsaturated fatty acids) is essential for optimal brain functioning (for an excellent review, see Gómez-Pinilla, 2008).

Conversely, a diet low in essential nutrients, and high in saturated fat and sucrose, has been shown to reduce hippocampal brain-derived neurotrophic factor (BDNF) and, as a result, hamper cognitive performance and synaptic plasticity (Molteni et al., 2002). Suboptimal cognitive functions, in turn, appear to predispose children to poor self-regulation (Jackson & Beaver, 2013) and related behavioral problems (Espy, Sheffield, Wiebe, Clark, & Moehr, 2011; Riggs et al.,

2004; Schoemaker et al., 2013).

Nutritional factors may be especially important to cognitive and behavioral development during the first months and years of life, as the brain is both a) experiencing exponential growth during this time and b) placing increasing demands on exogenous nutrients to supply the building blocks (e.g., proteins) that facilitate such growth (see Benton, 2008; Georgieff, 2007). Although a large number of studies have explored the relevance of childhood nutritional factors to healthy

21 brain development (Taki et al., 2010; for reviews, see Bellisle, 2004; Ryan et al., 2010), relatively few studies have explored whether nutrition during infancy and early childhood directly contributes to the development of externalizing behavior problems (Galler et al., 2011;

Liu et al., 2004). Importantly, animal research suggests that deprivation of key nutrients (e.g., omega-3 fatty acids) at critical developmental periods not only reduces synaptic differentiation and formation, but also increases aggressive behavior by disadvantageously altering serotonin levels (Hibbeln, Ferguson, & Blasbalg, 2006). The limited number of studies with human subjects suggests that similar neurological and behavioral outcomes may occur as a result of poor nutrition (Galler et al., 2011; Liu et al., 2004; Sinn, 2008).

3.1 Infant Nutrition

Importantly, the content and form of nutrition varies substantially across the first few years of life. For example, during the first months of life, infants are incapable of ingesting solid foods, and therefore typically rely on breast milk and/or infant formula for their nourishment. Of these two forms of infant nutrition, breast milk has been touted by researchers and medical professionals alike as the ideal form of infant sustenance, primarily due to the unique nutritional content of breast milk (see Ballard & Morrow, 2013; Lönnerdal, 2013). Specifically, breast milk is distinct from infant formula in that is contains a unique amalgam of bioactive proteins, nutrients, and growth factors that facilitates infant development (Newton, 2004).

Long-chain polyunsaturated fatty acids, such as docosahexaenoic acid (DHA) and arachidonic acid (AA), have been given the most empirical attention, due to their important role in promoting the structural and functional integrity of the brain (Singh, 2005). Despite the supplementation of many infant formulas with DHA and AA in recent years, infant formula has

22 not become the nutritional equivalent of breast milk, largely because breast milk contains a host of proteins and growth factors (e.g., lactoferrin, alpha-lactalbumin, milk fat globule membranes, brain-derived neurotrophic factor, glial cell lined-derived neurotrophic factor) that are still not present, or are only minimally present, in modern infant formula (Ballard & Morrow, 2013;

Lönnerdal, 2013; Newton, 2004). Some scholars have even suggested that certain components of breast milk not currently found in infant formula (e.g., bile salt-stimulated lipase) can enhance the absorption of long-chain polyunsaturated fatty acids (Hernell, Bläckberg, Chen, Sternby, &

Nilsson, 1993; Wright, Coverston, Tiedeman, & Abegglen, 2006; see also Ballard & Morrow,

2013), which may explain why a recent meta-analysis found no evidence of enhanced cognition among infants fed with DHA-enhanced formula (Qawasmi, Landeros-Weisenberger, Leckman,

& Bloch, 2012). Importantly, the heating process, as well as the low digestibility of the bovine corollaries of various human milk proteins, can also diminish the nutritional advantages of enhanced formula (Lönnerdal, 2013).

The benefits of the various human milk proteins are still not fully understood, but research to date suggests that, individually and/or collectively, they enhance nutrient absorption, stimulate physical growth, defend against pathogens, cultivate gut microorganisms, and promote neuronal survival, growth and plasticity (Ballard & Morrow, 2013; Nassar, Younis, El Arab, &

Fawzi, 2011). Furthermore, the composition of human milk is dynamic in that it varies‐ from feeding to feeding, from day to day, from mother to mother, from population to population, and across stages of lactation (Ballard & Morrow, 2013). Ultimately, this variability in milk composition has led some scholars to proclaim that “there is no real understanding of how to exactly mimic human milk” (Stam, Sauer, & Boehm, 2013, p. 526S). The tailor-made content of breast milk has been illustrated by research indicating that, for the first few weeks after birth, the

23 breast milk of mothers who give birth prematurely has a higher concentration of essential proteins and growth factors than the breast milk of mothers who give birth to full-term babies

(see Castellote et al., 2011; Dvorak, Fituch, Williams, Hurst, & Schanler, 2003). Infant formula, on the other hand, is standardized, which may reduce its ability to optimally meet the unique nutritional needs of infants at various stages of development.

As expected, research on the cognitive benefits of breastfeeding has, for the most part, been supportive (Belfort et al., 2013; Julvez et al., 2014; Lucas, Morley, Cole, Lister, & Leeson-

Payne, 1992; Quigley et al., 2012). For instance, in a seminal experiment by Lucas and colleagues (1992), preterm infants who were fed breast milk through a nasogastric tube (as opposed to formula) exhibited higher developmental scores at 18 months and an 8.3-point advantage in IQ at 7-8 years of age. A large, randomized trial of a breastfeeding promotion intervention across 31 Belarussian hospitals yielded similar results. Children who were still breastfeeding at 3 months of age exhibited higher verbal, performance, and full-scale IQ at age

6.5 than children who were weaned before 3 months of age (Kramer et al., 2008a). A more recent study, moreover, found that infants who breastfed for 6 months or more exhibit enhanced neurological functions (general, verbal, quantitative, memory, perception, and motor) at age 4

(Julvez et al., 2014). Interestingly, this effect was not accounted for by various familial factors

(e.g., maternal education, IQ, and social class) or by the concentration of omega-3 fatty acids present in colostrum, suggesting that the benefits of breast milk are not merely a product of omega-3 fatty acids.

In comparison to the plethora of studies exploring the link between breastfeeding duration to offspring cognition, relatively few studies have been conducted examining the influence of breastfeeding on early childhood behavioral problems. The studies to date,

24 however, are also largely supportive (Heikkilä et al., 2011; Julvez et al., 2007; Mimuoni-Bloch et al., 2013; Sabuncuoglu, Orengul, Bikmazer, & Yilmaz Kaynar, 2014), despite some null results

(Kramer et al., 2008b; Lind, Li, Perrine, & Schieve, 2014). Overall, the research suggests that long-term breastfeeding (typically defined as >= 4 or 6 months) is associated with lower odds of behavioral problems and diagnoses during early childhood (Heikkilä et al., 2011; Mimuoni-

Bloch et al., 2013; Sabuncuoglu et al., 2014). To illustrate, a study by Julvez and colleagues

(2007) revealed that breastfeeding long term decreases the risk of offspring behavioral problems related to attention, hyperactivity, and social competence at age 4. Some research has found evidence that the influence of breastfeeding may even extend into adolescence and affect various dimensions of mental health (Oddy et al., 2010).

3.2 Early Childhood Nutrition

Of course, as children grow, they begin transitioning from a diet of formula and/or breast milk to a solid-foods diet, so that by early childhood, their nutritional repertoire can potentially involve a number of different foods, from carrots and celery to cookies and chocolate cake.

Thus, in addition to breastfeeding during infancy, the quality of nutrition during early childhood might also influence both cognitive and behavioral development. Nutritious diets at this stage of development can contribute to enhanced intelligence and cognitive functions as children age

(Benton, 2008; Nyaradi, Li, Hickling, Foster, & Oddy, 2013; Ryan et al., 2010; Taki et al. 2010;

Theodore et al., 2009), whereas less nutritious diets appear to impede brain development

(Northstone, Joinson, Emmett, Ness, & Paus, 2011; Riggs, Spruijt-Metz, Sakuma, Chou, &

Pentz, 2010). Theodore and colleagues (2009), for instance, found that children who ate fish weekly at age 3.5 exhibited higher IQ scores at age 7 than children who ate fish less frequently.

25 Conversely, a study by Northstone and colleagues (2011) revealed that a diet high in fat, sugar, and processed foods at ages 3, 4, and 7 predicted significant reductions in IQ by age 8.

A number of studies have also suggested that poor nutrition during early childhood can contribute to the development of behavioral problems (Galler et al., 2011; Liu et al., 2004; Oh et al., 2013; Park et al., 2012; Wiles et al., 2007; Woo et al., 2014). To illustrate, a study by Liu and colleagues (2004) found that children who exhibited signs of acute malnutrition during the first few years of life engaged in more externalizing behaviors at ages 8, 11, and 17, implying that nutritional factors during the earliest stages of the life course may indeed play a role in the development of aggressive and hyperactive behavior. Research has also indicated that a western diet, defined as a diet high in “junk food”, can contribute to behavioral problems in children

(Wiles et al., 2007; Woo et al., 2014).

In sum, there is a substantial amount of evidence that various nutritional factors during infancy and early childhood can contribute to the cognitive and behavioral development of children. Breastfeeding for a longer duration appears to enhance neuropsychological functions and minimize externalizing behavior through early childhood (and potentially beyond; see Oddy et al., 2010). Furthermore, malnutrition and poor diet quality during early childhood are associated with subsequent behavioral problems and deficits in intelligence (Liu et al., 2004;

Northstone et al., 2011; Woo et al., 2014). The research generally supports the notion, therefore, that nutritional factors during the early stages of the life have significant implications for the development of externalizing behavior, a key risk factor for later crime and delinquency.

26 3.3 The Association between Nutritional Factors and Attachment Security

In addition to the research linking early life nutrition to both cognition and behavior

(Belfort et al., 2013; Heikkilä et al., 2011; Julvez et al., 2007; Oddy et al., 2010), some studies have shown a link between infant nutritional patterns and the attachment style of offspring during toddlerhood and/or childhood (Britton et al., 2006; Gribble, 2006; Tharner et al., 2012).

This line of research intimates that breastfeeding may assist in the development of a secure attachment between parent and child (Britton et al., 2006; Gribble, 2006; Tharner et al., 2012), which is reflected in the child’s ability to use their parent as “a secure base from which to explore the environment” (Britton, et al., 2006, p. 1437). This essential component of socioemotional development is facilitated by a process in which the child’s safety and survival is maintained through reciprocal parent-infant interactions that are characterized by closeness, sensitivity, and responsiveness. This process aids in the development of “better organized neural control mechanisms”, which are expected to result in “greater self-regulatory capacities” and, ultimately, an internalized working model of the self and the self in relation to others

(Andreassen & Fletcher, 2007, pp. 8.1-8.2; Bowlby, 1969). To date, research has suggested that several factors may help or hinder a child’s ability to form a secure attachment, including physical and sexual abuse (Lyons-Ruth, Connell, Zoll, & Stahl, 1987; Shapiro & Levendosky,

1999), child temperament (Planalp & Braungart-Rieker, 2013; Susman-Stillman, Kalkoske,

Egeland, & Waldman, 1996), marital discord (Frosch, Mangelsdorf, & McHale, 2000; Owen &

Cox, 1997), parental sensitivity (Bakermans-Kranenburg, Van Ijzendoorn, & Juffer, 2003; Wolff

& Ijzendoorn, 1997), and genetic factors (Lakatos et al., 2000; Raby et al., 2012). Still, a relatively small number of studies have explicitly explored whether breastfeeding during the first

27 months of life enhances attachment security (Britton et al., 2006; Gribble, 2006; Tharner et al.,

2012).

While it is possible that breastfeeding enhances attachment through heightened levels of support/comfort, it is also possible that the feeding method itself may have little to do with the security of attachment, or that it may only indirectly affect attachment security by enhancing the quality and frequency of positive dyadic interactions between the mother and her child.

Research to date on the topic is rather equivocal, with some studies findings significant relationships between breastfeeding and attachment security (Britton et al., 2006; Gribble, 2006;

Tharner et al., 2012), and others finding little to no evidence of a relationship (Else-Quest, Hyde,

& Clark, 2003; Papp, 2013; Jansen, Weerth, & Riksen-Walraven, 2008; Schulze & Carlisle,

2010). Still, in the face of somewhat inconclusive results, the rationale behind a breastfeeding- attachment link remains quite sound. As Gribble so eloquently stated, “Breastfeeding may assist attachment development via the provision of regular intimate interaction between mother and child; the calming, relaxing and analgesic impact of breastfeeding on children; and the stress relieving and maternal sensitivity promoting influence of breastfeeding on mothers” (p. 1).

There is some evidence to suggest, moreover, that breastfeeding increases the maternal oxytocin response (Feldman, Gordon, & Zagoory-Sharon, 2011; Nissen et al., 1996; Strathearn, Iyengar,

Fonagy, & Kim, 2012) as well as maternal sensitivity to the cues of offspring (Kim et al., 2011;

Tharner et al., 2012), both of which have been linked to attachment security (Leerkes, 2011;

Nelson & Panksepp, 1998; Pederson et al., 1990; Strathearn, Fonagy, Amico, & Montague,

2009).

If breastfeeding does indeed enhance attachment security, then it may exert an indirect effect on various child outcomes, since high attachment security has been found to have a

28 positive impact on numerous dimensions of children’s cognitive (Jacobsen, Edelstein, &

Hofmann, 1994; West, Mathews, & Kerns, 2013), psychological (Brumariu, Kerns, & Seibert,

2012; McLaughlin, Zeanah, Fox, & Nelson, 2012), socioemotional (Barone & Lionetti, 2012;

Boldt, Kochanska, Yoon, & Koenig Nordling, 2014; Panfile & Laible, 2012), and behavioral

(Goldner & Scharf, 2013; Torres, Maia, Veríssimo, Fernandes, & Silva, 2012) development.

Conversely, children who develop insecure attachment styles (e.g., disorganized, ambivalent, avoidant) appear to be at a higher risk of various negative outcomes across the life course, including anxiety (Brumariu et al., 2012; Colonnesi et al., 2011; Groh, Roisman, van IJzendoorn,

Bakermans Kranenburg, & Fearon, 2012), depression (Kamkar, Doyle, & Markiewicz, 2012;

Kerns, Brumariu,‐ & Seibert, 2011), ADHD (Storebo, Rasmussen, & Simonsen, 2013), externalizing behavior (Dubois-Comtois, Moss, Cyr, & Pascuzzo, 2013; Groh et al., 2012;

Kochanska & Kim, 2013; O’Connor et al., 2012), and even delinquency (Allen et al., 2002).

Therefore, to the extent that breastfeeding increases attachment security, it is possible that infants with little to no exposure to breastfeeding may develop an insecure attachment to caregivers and, as a result, be at risk of a host of negative outcomes, including externalizing behavior.

It is also possible that a secure attachment developed during the first few years of life may set the stage for beneficial nutritional choices during childhood. Although research on this topic is somewhat limited, studies to date suggest a possible link between attachment security and subsequent eating behaviors of offspring during childhood (Bost et al., 2014; Bozorgi,

Sho’leh Amiri, & Talebi, 2014; Faber & Dubé, 2015; Simmons, Goldberg, Washington, Fischer-

Fay, & Maclusky, 1995; for an excellent summary, see Lu, Faber, & Dubé, 2013) and later life stages (Aarts et al., 2015; Huntsinger & Luecken, 2004). Deficiencies in micronutrients, such as iron, have also been associated with poor mother-child reciprocity (e.g., turn taking, shared

29 positive affect, eye contact, etc.) (see Corapci, Radan, & Lozoff, 2006). While possible mechanisms are numerous, scholars generally posit that an insecure attachment between the caregiver and the child result in poor emotional regulation of the child, which increases his/her unhealthy eating behaviors by influencing caregiver feeding practices as well as the child’s temperament/irritability (see Bost et al., 2014, for a hypothetical causal model). Importantly, additional studies have found that the formation of an insecure attachment during infancy and toddlerhood increases the risk of obesity (Anderson, Gooze, Lemeshow, & Whitaker, 2012;

Anderson & Whitaker, 2011; Trombini et al., 2003), which may be partly explained by poor eating habits. Anderson and colleagues (2012) also argue that attachment security influences the child’s ability to regulate negative affect and stress responses, which can interfere with metabolic functioning and increase the frequency of sweet and salty food intake (see also Bost et al., 2014).

As noted previously, poor eating habits have been found to increase the risk of externalizing behavior (Howard et al., 2011; Wiles et al., 2007; Woo et al., 2014). In light of the body of findings linking dietary choices to attachment security, it is reasonable to suggest that the relationship between the eating habits and the externalizing behavior of children may be explained, at least in part, by attachment security. In short, since poor diet during childhood appears to be associated with low attachment security, and low attachment security has repeatedly been linked to externalizing behavior, it may be that dietary behaviors have little effect on externalizing behaviors independent of their relationship with attachment security.

Attachment security during the first few years of life, therefore, may play a pivotal role in the link between nutritional factors and externalizing behavior during childhood.

30 3.4 Key Limitations in the Nutrition Literature

Although research to date is mostly supportive of the link between nutritional factors during infancy and childhood and the development of externalizing behavior, most of this research has been conducted outside of the field of criminology. Moreover, criminologists have not adequately incorporated the findings from research on infant and early childhood nutrition into mainstream criminological theory and research. In addition to this key oversight, the current literature presently suffers from five important limitations. These limitations are both methodological and theoretical in nature.

3.4.1 Familial Confounding

The first of these limitations is familial confounding, or selection bias stemming from factors within the family/home environment. This shortcoming is rooted in the frequent use of observational data that are based on samples of only one child per household (for examples, see

Heikkilä et al., 2011; Julvez et al., 2007; Lind et al., 2014; Oddy et al., 2010; Park et al., 2012;

Sabuncuoglu et al., 2014; Woo et al., 2014). Results derived from these data are subject to familial confounding, since environmental risks of interest (e.g., poor nutrition) can be correlated with (and confounded by) other factors within the family environment (e.g., socioeconomic status, maternal education, maternal personality). When researchers examine only one child per household, they are unable to identify the degree to which an environmental influence such as nutrition is shared by siblings within that household. Thus, in observational studies that use one child per household, the inclusion of statistical controls for potential confounding influences within the family environment is a common strategy to increase the internal validity of the findings.

31 However, there is growing concern among scholars that, even after including statistical controls for important confounders in the family environment, residual confounding may render the relationship between nutritional factors (e.g., breastfeeding) and externalizing behavior spurious, due to omitted variable bias and/or poor measurement (Colen & Ramey, 2014;

Evenhouse & Reiley, 2005; Kramer et al., 2008b; Kramer et al., 2011). These concerns are rooted in the divergent results of a handful of studies that use alternative methodological tools that more effectively address the issue of selection bias due to familial factors (see Colen &

Ramey, 2014; Evenhouse & Reiley, 2005; Kramer et al., 2008b). For example, while most observational studies yield moderate support for the role of breastfeeding duration in the development of externalizing behavior (although see Lind et al., 2014), randomized control trials and sibling designs are much less supportive of the link between breastfeeding and externalizing behavior (Colen & Ramey, 2014; Evenhouse & Reiley, 2005; Kramer et al., 2008b; Kramer et al., 2011).

Kramer and colleagues (2008b), for instance, conducted a cluster-randomized trial of a breastfeeding intervention program using a sample of nearly 14,000 mother-child dyads in the republic of Belarus. At the 6.5 year follow-up, there was no evidence that prolonged breastfeeding resulted in improved affect or behavior of offspring during early childhood. A recent sibling study by Colen and Ramey (2014), furthermore, found that the relationship between breastfeeding duration, child hyperactivity, and behavioral non-compliance was not significant once siblings were used to neutralize confounders within the family environment that impact siblings symmetrically. Thus, while between-family comparisons seem to generally support the link between breastfeeding and offspring externalizing, the limited research using within-family and experimental designs is typically not supportive. The discordant results

32 suggest that residual confounding, or the inability to appropriately control for selection bias, may be resulting in a body of literature that is unduly and erroneously supportive of the breastfeeding- externalizing link (see also Kramer et al., 2011; D’Onofrio et al., 2013 for more discussion on residual confounding). In a similar manner, familial processes (e.g., socioeconomic status, maternal education, maternal personality) may confound the link between early childhood nutrition (e.g., “junk food” intake) and externalizing behavior, but scholars have yet to fully account for this possibility using a sibling comparison design.

3.3.2 Genetic Confounding

A significant, though rarely noted limitation of the literature examining the link between nutritional factors and childhood behavioral problems is its lack of attention to genetic confounding. Specifically, whether examining breastfeeding or early childhood nutrition, selection into nutritious or non-nutritious environments appears to be correlated with and/or influenced by genetic factors, in addition to family environmental factors (see Colodro-Conde,

Sánchez-Romera, & Ordoñana, 2013; Colodro-Conde et al., 2014; Faith et al. 2006). Genetic factors, in turn, have been shown to play a significant role in the development of externalizing behavior (Baker et al., 2007; Barnes et al., 2013). Therefore, if similar genetic risk factors underpin both exposure to inadequate nutrition during infancy and childhood as well as the development of externalizing behavior, then the link between nutrition and externalizing may be spurious due to genetic factors. Similar arguments have been made concerning the potential for genetic factors to confound the association between externalizing behaviors in children and other early-life risk factors relating to the mother (e.g., prenatal maternal smoking and alcohol use)

(Boutwell & Beaver, 2010; D’Onofrio et al., 2008; Knopik et al., 2006; Knopik, 2009; Thapar et al., 2009). Although many of these studies have been unable to explicitly model genetic risk (see

33 Boutwell & Beaver, 2010), studies that are able to model genetic risk suggest that genetic factors may explain the association between externalizing behavior and early-life risk factors associated with the mother and/or family environment (see Knopik et al., 2006; Thapar et al., 2009).

In the case of breastfeeding, it is possible that shared genetic factors between the mother and her child underpin both the mother’s propensity to initiate and continue breastfeeding as well as the child’s tendency toward externalizing behaviors, as mothers and their biological children share 50% of their distinguishing DNA. Thus, genetically influenced traits, such as low self- control (Beaver et al., 2009), might a) be shared by mother and child due to shared genetic material, b) reduce the likelihood of successful, prolonged breastfeeding, and c) increase the likelihood of behavioral problems in the child. In this example, the mother would transmit both genetic risk and environmental risk to the child, and these risks are likely correlated, making it difficult to tease out the independent effects of genes and the environment on externalizing behavior.

To date, only one study of the link between infant nutrition and childhood conduct problems has accounted for genetic confounding (Shelton et al., 2011). The authors employed a unique cross-fostering design to compare mother-child dyads who were genetically related to those who were not genetically related. Upon examining the behavior of children who were breastfed by a mother with whom they shared no genetic material (i.e., due to egg or embryo implantation), Shelton and colleagues (2011) found no evidence of a relationship between breastfeeding duration and offspring conduct problems. Although replication is needed, the results of the study suggest that genetic confounding of the link between breastfeeding and offspring conduct problems is a potentially important threat to this body of literature, and that

34 researchers who do not employ genetically informative designs may draw erroneous conclusions about the causal influence of early-life nutrition on behavioral problems.

In the case of early childhood nutrition, it is possible that children may exhibit both poor eating habits and poor behavior as a result of a genetically influenced latent trait. Thus, the challenging eating behaviors and the challenging social behaviors may stem from a similar source (e.g., low self-control, difficult temperament) that might, at least in part, have genetic underpinnings (Beaver et al., 2009; Wright, Beaver, DeLisi, & Vaughn, 2008). To the extent that this process is occurring, studies of early childhood nutrition, and its effect on externalizing behavior, might be misspecified. No research to date, however, has explicitly examined whether the relationship between early childhood diet and externalizing behavior is robust to the influence of genetic factors.

3.3.3 Low Attachment Security as an Intervening and/or Antecedent Process

The third limitation of the literature is the absence of research considering the role of low attachment security in the relationship between nutritional factors and externalizing behavior.

As noted previously, both breastfeeding practices and externalizing behavior have been independently linked to attachment security (Britton et al., 2006; Guttmann-Steinmetz &

Crowell, 2006; Kochanska & Kim, 2013; O’Connor et al., 2012; Tharner et al., 2012), despite some contradictory results (Schulze & Carslile, 2010; Papp, 2013). Still, researchers have yet to fully explore the extent to which attachment security during toddlerhood mediates the link between breastfeeding during infancy and externalizing behavior during early childhood. Most scholars do not even statistically adjust for attachment security when calculating their results

(although see Heikkilä et al., 2011, for an exception). Oddy and colleagues (2010), for instance, detected significant associations between breastfeeding duration and subsequent externalizing

35 and mental health problems, but did not explore the role of attachment security in their findings.

Still, they acknowledged that “breastfeeding may be an indicator of secure attachment status, which is known to have a positive influence on [a] child’s psychological development” (p. 573).

More recently, Hayatbakhsh, O’Callaghan, Bor, Williams, and Najman (2012) noted the “need to investigate whether the child-mother relationship … explain[s] the association between breastfeeding and the offspring’s [extent of] social, attention, and aggression problems” (p. 485).

Such empirical examinations are clearly essential in order to narrow in on an explanation of the association between breastfeeding and externalizing behavior.

Low attachment security might also impact externalizing behavior through its influence on childhood eating habits (Lu et al., 2013). As noted previously, researchers have found some evidence suggesting that an insecure attachment increases the risk of poor eating habits (Bost et al., 2014; Bozorgi et al., 2014; Faber & Dubé, 2015; Goossens, Braet, Bosmans, & Decaluwé,

2011). Poor eating habits, in turn, have been found to increase the risk of behavioral problems in children (Wiles et al., 2007; Woo et al., 2014). Whether childhood eating habits operate as a mediating pathway linking attachment security to externalizing behavior, however, has yet to be examined. This omission in the literature is rather surprising, since the relationship between insecure attachment and externalizing is frequently explained by an important correlate of poor nutritional intake during childhood – negative emotional regulation (see Anderson et al., 2012;

Bost et al., 2014; Braet & van Strien, 1997; Guttmann-Steinmetz & Crowell, 2006). It is thus reasonable to suggest that childhood nutritional factors may partly explain why an insecure attachment increases the risk of subsequent externalizing behavior.

36 3.3.4 Nutrition as a Biosocial Process

The fourth limitation of the literature is the minimal attention given to the role of biosocial processes (e.g., gene-environment interactions, or GxE). Only one study to date, for example, has considered whether the influence of infant or early childhood nutrition on externalizing behavior is significantly different for individuals with varying degrees of genetic risk (Groen-Blokhuis et al., 2013). It is possible, for instance, that breastfeeding might be differentially protective for children depending on their genetic make-up. One polymorphism that has been examined as a potentially important moderator of the breastfeeding-cognition relationship is FADS2 (Caspi et al., 2007; Morales et al., 2011; Rizzi et al., 2013). This gene is implicated in the control of fatty acid pathways in the brain. Because breast milk is known to have unique fatty acid content, and FADS2 is implicated in the metabolism of fatty acids, researchers have hypothesized that the benefits of breastfeeding might be conditioned by

FADS2. To date, results have been mixed, with some studies finding evidence of GxE (Caspi et al., 2007; Morales et al., 2011; Rizzi et al., 2013), and other studies finding little to no evidence of GxE (Groen-Blokhuis et al., 2013; Martin et al., 2011; Steer, Smith, Emmett, Hibbeln, &

Golding, 2010). Despite these empirical examinations of the role of nutrient-gene interactions in the development of cognition, few studies have explored the role of nutrient-gene interactions in the development of childhood behavioral problems. In a notable exception, Groen-Blokhuis and colleagues (2013) tested for the presence of an interaction between breastfeeding and FADS2 in the prediction of overactivity at age 3 and attention problems at ages 7, 10, and 12. While there was some evidence of small, direct effects of breastfeeding and two FADS2 single nucleotide polymorphisms (rs174575 and rs1535), there was no evidence of a breastfeeding x FADS2 interaction in the prediction of overactivity/attention problems.

37 Virtually none of the empirical attention has been given to genetic factors as moderators of the relationship between childhood diet and externalizing behavior (although, see Field, 2014 for a discussion). Instead, the literature typically links gene-nutrient interplay to the cognition

(Dauncey, 2012; Mattson, 2003; Whalley et al., 2008) and physical health (Karnehed, Tynelius,

Heitman, & Rasmussen, 2006; Mathers & Hesketh, 2007; Ordovas, 2006) of older subjects, generally neglecting the potential for genes and diet to interact in their prediction of childhood behavioral patterns. Studies exploring the moderating role of genetic factors in the nutrition- externalizing relationship are clearly needed, since research has repeatedly indicated that genetic factors can make children differentially susceptible to early environmental influences (Belsky &

Beaver, 2011; Jaffee et al., 2005).

Some research has suggested that children may be especially susceptible to poor nutrition if they are born low birth weight (Quigley et al., 2012; Smith, Durkin, Hinton, Bellinger, &

Kuhn, 2003). For example, medical research has indicated that, during the first 4 weeks following birth, the breast milk (including colostrum) of mothers who give birth to premature, low-birth-weight infants is characterized by higher protein, energy, and lipid content than the breast milk of mothers who give birth to full-term infants (Atkinson, Bryan, & Anderson, 1981).

Low birth weight infants who never initiate breastfeeding, or who are breastfed for a shorter duration, may therefore be especially vulnerable to poor health, and possibly behavioral, outcomes. Moreover, research has indicated that low birth weight children are at greater risk of behavioral problems, particularly ADHD symptomatology, than normal birth weight children

(Bhutta, Cleves, Casey, Cradock, & Anand, 2002; Botting, Powls, Cooke, & Marlow, 1997;

Jackson & Beaver, 2015; Mick et al., 2002). Research has also revealed that breastfeeding significantly improves cognitive functioning in preterm infants, at least in part, by increasing

38 brain DHA levels (Tanaka, Kon, Ohkawa, Yoshikawa, & Shimizu, 2009). Specifically, a meta- analysis by Anderson, Johnstone, and Remley (1999) revealed that the cognitive development of premature infants is influenced by breast milk to a significantly greater degree than the cognitive development of full-term infants (see also Quigley et al., 2012). It is reasonable to suggest, therefore, that inadequate nutrition during infancy might exacerbate externalizing behavior problems among low birth weight children. Infancy likely represents a more critical point of intervention for the cognitive development of preterm or low-birth-weight infants. For instance, adequate nutrition during the first months of life might permit low birth weight infants to developmentally “catch up” to normal birth weight infants (see Cooke, 2006, for an excellent discussion). Still, no research to date has examined whether birth weight moderates the breastfeeding-externalizing relationship.

Finally, it is possible that inadequate nutrition during infancy and early childhood may impact the externalizing behavior of children differently depending on their level of attachment security. For example, a shorter duration of breastfeeding (or failure to initiate breastfeeding) may have little impact on externalizing behavior for those children who are able to form a secure attachment to their caregiver, despite minimal or no exposure to breastfeeding. In other words, while breastfeeding may aid the development of a secure attachment between infant and mother, children who have little to no exposure to breastfeeding may not be at great risk of externalizing behavior if they are able to achieve attachment security through other means (i.e., in the absence of breastfeeding). Scholars have yet to examine this possibility, as the literature on breastfeeding and attachment has largely examined direct causal paths between the two variables (Britton et al.,

2006; Tharner et al., 2012). However, a recent study by Liu, Leung, and Yang (2013) revealed that breastfeeding may protect against internalizing behaviors, but only in the context of

39 adequate mother-child bonding. The authors argue that their findings highlight the presence of a

“biosocial and holistic effect of physiological, nutritive, and maternal-infant bonding” (p. 76).

While the study did not examine externalizing behaviors specifically, it suggests that breastfeeding may be differentially effective on the basis of the quality and/or extent of mother- infant attachment. Specifying the socioemotional factors that might condition the influence of breastfeeding on externalizing behavior is an important next step in this body of research, particularly since the findings examining the direct effects of breastfeeding on externalizing are somewhat inconclusive (Julvez et al., 2007; Kramer et al., 2008b; Lind et al., 2014; Oddy et al.,

2010).

Nutritional factors during early childhood might also influence externalizing behaviors differentially depending on the attachment style of the child. A securely attached child, for example, may be less susceptible to the harmful effects of a poor diet on externalizing behavior, as the emotional/social support and enhanced coping capacities that characterize their attachment style may indeed counteract the risks incurred from consuming a low quality diet. Conversely, children who a) are insecurely attached to their caregiver and b) ingest a diet low in nutrient- dense foods may be especially at risk of exhibiting externalizing behavior, as they would be exposed to compounding risk factors that are likely to result in social, emotional, and cognitive deficits. No research to date has examined these possibilities. Randomized control trials have suggested that children’s behavior can indeed improve as a result of dietary supplementation of key nutrients such as omega-3 fatty acids, zinc, and magnesium (Schoenthaler & Bier, 2000;

Sinn & Bryan, 2007). It is reasonable to suggest, however, that the size of this effect may differ depending on whether the child is securely attached to their caregiver. Children who are already equipped with the ability to effectively regulate their emotions and cope with challenges may not

40 experience many changes in behavior as a result of dietary changes. Conversely, their socioemotional connection with their caregiver might make them especially responsive to dietary treatment. Additional research is clearly needed to determine whether attachment security moderates the nutrition-externalizing relationship.

3.3.5 The Interplay between Nutritional Factors across Infancy and Early Childhood

The fifth and final limitation of the literature is the seemingly unwarranted disconnect between the bodies of research that examine a) nutritional factors during infancy and b) nutritional factors during childhood, particularly as these bodies of research pertain to childhood antisocial behavior. Breastfeeding, as a form of infant nutrition, is typically studied in isolation from dietary behaviors at subsequent stages of the life course (however, see Perrine et al., 2014).

Even though breastfeeding has been linked to a host of health outcomes, including body weight

(Armstrong & Reilly, 2002; Dewey, 2003; Harder, Bergmann, Kallischnigg, & Plagemann,

2005), immunologic health (Jackson & Nazar, 2006), and blood pressure (Martin, Gunnell, &

Smith, 2005), scholars have generally neglected to explore the possibility that little to no exposure to breastfeeding may increase the risk of poor eating habits during childhood, which may explain its effect on outcomes like obesity (Armstrong & Reilly, 2002; Dewey, 2003).

Although Dewey (2003) noted that the effect of breastfeeding on obesity, for example, is likely explained in part by “learned self-regulation of energy intake”, she never explicitly examined the effect of breastfeeding on childhood eating habits (p. 9). A recent study by Perrine and colleagues (2014), however, provided some evidence that exposure to breastfeeding may increase the likelihood of developing healthier eating habits during childhood (see also Abraham,

Godwin, Sherriff, & Armstrong, 2012; Grieger, Scott, & Cobiac, 2011; Scott et al., 2012). Of course, if breastfeeding does indeed predict childhood dietary behaviors, then it may also

41 indirectly impact externalizing behavior by setting in motion a pattern of nutrition that may heighten or reduce the risk of conduct problems during childhood. Research to date has yet to explore this possibility.

Although a handful of scholars have examined the potential for breastfeeding to improve subsequent eating habits, the question of whether breastfeeding and childhood diet interact to predict externalizing behavior has been completely overlooked. For example, it is possible that breastfed infants may be less susceptible to the negative effects of poor dietary choices at later life stages, due to various proteins, growth factors, and microbiota present in breast milk that impact metabolism (see Dieterich, Felice, O’Sullivan, & Rasmussen, 2013; Lönnerdal, 2013).

Infants who are not breastfed are not exposed to the unique amalgam of proteins and growth factors contained in breast milk, including BDNF, which supports brain growth and plasticity, and minimizes neurotoxicity. As a result, these infants may be more vulnerable to the potentially deleterious effects of poor dietary habits during childhood on cognition and behavior. Evidently, it is difficult to predict exactly how the health-promoting properties of breast milk would interact with subsequent diet to predict behavioral problems. Even so, since both breastfeeding and diet quality have been found to impact the likelihood of externalizing behavior, it is clearly possible that nutritional risk factors across distinct stages of development may accumulate to increase the likelihood of externalizing behavior.

3.5 Contribution of the Current Study

This dissertation seeks to address the aforementioned voids in the literature by employing a genetically informative, twin-based research design to examine whether twin differences in exposure to nutritional factors during infancy and early childhood are predictive of twin

42 differences in externalizing behavioral problems during kindergarten. The design allows me to more effectively address the issues of familial and genetic confounding that so frequently plague this body of research. It also permits an examination of the interactive role of nutritional and genetic factors in the prediction of externalizing behavior as well as the role of other biosocial processes (e.g., breastfeeding x low birth weight) in the prediction of externalizing behavior.

Additionally, the current research design enables me to more rigorously explore the role of attachment security in the relationship between nutritional factors and externalizing behavior.

In general, this dissertation seeks to answer the call of scholars such as Moffitt (2005),

Jaffee et al. (2012), and D’Onofrio et al. (2013), who have acknowledged the need for sibling designs and other statistical innovations in order to a) bridge the gap between the biological and the social sciences and b) more effectively distinguish between environmental correlates and causes of antisocial behavior. In short, the current study uses a genetically informative sibling design in order to tease apart the independent, indirect, and interactive effects of nutritional, genetic, socioemotional and perinatal factors on the development of externalizing behavior.

Specifically, the present study will address the following eleven questions:

Infant Nutrition:

1. In what ways do mothers and children who breastfeed for shorter durations differ from

mothers and children who breastfeed for longer durations?

2. Does a short duration of breastfeeding significantly increase the risk of low attachment

security during toddlerhood and/or externalizing behavior problems during kindergarten,

independent of familial and genetic influences?

43 3. Is the relationship between short duration of breastfeeding and externalizing behavior

problems during kindergarten explained by low attachment security during toddlerhood

and/or low birth weight?

4. Is the influence of breastfeeding duration on kindergarten externalizing behavior

moderated by genetic risk, low attachment security, and/or low birth weight?

Early Childhood Nutrition:

5. In what ways do children with especially poor dietary habits differ from other children?

6. Does a low quality diet during preschool significantly increase the risk of externalizing

behavior problems during kindergarten, independent of familial and genetic influences?

7. Is the relationship between a low quality diet during preschool and externalizing behavior

during kindergarten robust to indicators of low attachment security during toddlerhood

and externalizing behavior during preschool?

8. Is the influence of a low quality diet on kindergarten externalizing behavior moderated by

genetic risk and/or low attachment security?

Interplay Between Nutritional Factors:

9. Does a short duration of breastfeeding increase the likelihood of a low quality diet?

10. Is the relationship between short duration of breastfeeding and externalizing behavior

problems during kindergarten explained by a low quality diet during preschool?

11. Do a short duration of breastfeeding and a low quality diet during preschool interact to

predict kindergarten externalizing behavior?

44 CHAPTER 4

NUTRITION AND EXTERNALIZING BEHAVIOR:

A MULTIDISCIPLINARY REVIEW

A multidisciplinary review is essential in order to adequately cover the literature that is germane to this dissertation. Myriad disciplines are represented among the studies examining the relationship between infant and/or child nutrition and behavior, including medicine (Galler et al.,

2011), public health (Galler et al., 2011; Hayatbakhsh et al., 2012; Heikkilä et al., 2011), epidemiology (Julvez et al., 2007; Kramer et al.,2008b; Oddy et al., 2010), pediatric neurology

(Mimouni-Bloch et al., 2013), psychiatry (Park et al., 2012; Sabuncuoglu et al., 2014), and nutritional science (Oh et al., 2013), with criminology being represented only rarely (Liu et al.,

2004; Raine, 2002). Furthermore, due to the interdisciplinary, biosocial nature of the present research questions, and in order to properly situate the research questions within the current literature, a review of the relationship between externalizing behavior and genetic, perinatal, and socioemotional factors is also necessary.

This review will begin by examining the relationship between infant nutrition and externalizing behavior. Next, the literature on the link between attachment security, infant nutrition, and externalizing behavior will be discussed. Similar reviews will also be provided for studies examining early childhood nutrition. Next, the role of both genetic and perinatal factors in the development of childhood antisocial behavior will be reviewed. Following these sections, a rationale for possible biosocial interactions involving nutritional, genetic, perinatal, and socioemotional factors will be supplied by reviewing research that examines the biosocial origins of antisocial behavior. Finally, recent research exploring the interplay between nutritional factors across life stages will be reviewed, and gaps in this literature will be discussed.

45 4.1 Breastfeeding and Externalizing Behavior

As noted previously, the vast majority of research on the benefits of a longer duration of breastfeeding has examined health outcomes, including infant growth (Kramer et al., 2003), immunity (Hanson, 1998), child obesity (Brion et al., 2011), blood pressure (Hosaka et al., 2012) and cognition (Belfort et al., 2013; Boutwell, Beaver, & Barnes, 2012). Whether breastfeeding duration impacts offspring behavioral problems is less commonly examined (Hayatbakhsh et al.,

2012; Heikkilä et al., 2011; Julvez et al., 2007). The less frequent consideration of behavioral outcomes in this literature is somewhat surprising, since research has linked breastfeeding to neuropsychological functioning during early childhood (Julvez et al., 2007; Oddy et al., 2011), and poor neuropsychological functioning during early childhood has been found to predict misconduct later in life (Jackson & Beaver, 2013). Furthermore, various structural features of the brain have also been linked to breastfeeding duration, including brain size (Isaacs et al.,

2010), white matter density in prefrontal regions (Deoni et al., 2013; Isaacs et al., 2010), and cortical thickness of superior parietal lobules (Kafouri et al., 2012). Importantly, some of the brain features that have been associated with a longer duration of breastfeeding have been implicated in the degree of emotional and behavioral regulation. For example, abnormalities in the white matter density of prefrontal regions have been associated with ADHD symptomatology, including impulsivity, hyperactivity, and associated behavioral problems

(Helpern et al., 2011; Silk, Vance, Rinehart, Bradshaw, & Cunnington, 2009). Furthermore, recent research has also revealed that conduct-disordered adolescents exhibit diminished cortical thickness of both the right and left superior parietal lobules (Hyatt, Haney-Caron, & Stevens,

2012). In light of these findings, it is possible that brain areas impacted by the initiation and/or

46 duration of breastfeeding may be directly implicated in the likelihood of developing conduct problems during childhood and adolescence.

A number of researchers have hypothesized that breastfeeding duration may have important implications for the behavior and mental health of offspring as they age (Borra,

Iacovou, & Sevilla, 2012; Hayatbakhsh et al., 2012; Heikkilä et al., 2011; Julvez et al., 2007;

Kwok, Leung, & Schooling, 2013; Lind et al., 2014; Mimouni-Bloch et al., 2013; Oddy et al.,

2010; Oddy et al., 2011; Reynolds, Hennessy, & Polek, 2014; Sabuncuoglu et al., 2014; Shelton et al., 2011; Yorifuji et al., 2014). Overall, the research to date suggests that breastfeeding duration is associated with behavioral problems in offspring, with shorter durations increasing the risk of impulsive, hyperactive, and aggressive behaviors (Heikkilä et al., 2011; Julvez et al.,

2007; Mimouni-Bloch et al., 2013; Oddy et al., 2010; Sabuncuoglu et al., 2014; Yorifuji et al.,

2014). For instance, Sabuncuoglu and colleagues (2014) recently compared a Turkish sample of

200 children and adolescents diagnosed with ADHD with a control sample of 175 healthy schoolchildren. The results indicated that the children with ADHD were significantly more likely than controls to have experienced insufficient exclusive breastfeeding (defined as less than

6 months), a long duration of bottle-feeding, and an early initiation of bottle-feeding. A study by

Mimouni-Bloch and colleagues (2013) of a sample of Israeli children ages 6 to 12 found similar results. The study revealed that, while 73% of non-ADHD controls were still breastfeeding at three months of age, only 43% of children with ADHD were still breastfeeding at three months of age. The significant difference between groups also persisted at the 6-month mark, and was robust to maternal age at birth, infant gender, and parental divorce. One study of a Japanese

2001 birth cohort has even suggested that exclusive breastfeeding for a relatively long duration

47 (at least 6 to 7 months) is especially protective against hyperactive and impulsive behaviors during early childhood (Yorifuji et al., 2014).

In addition to its impact on ADHD symptomatology, research has indicated that breastfeeding duration may also influence other associated behavioral problems in offspring, including peer problems, inappropriate conduct, diminished prosocial behavior, and impaired social competence (Heikkilä et al., 2011; Julvez et al., 2007). To illustrate, Julvez and colleagues (2007) examined a sample of four-year-old Spanish children from two distinct prospective population cohorts. At age four, the children underwent professional evaluation by psychologists to determine their behavioral and neurological well-being. The results indicated that, in addition to lower attention and higher hyperactivity scores, children who were breastfed for less than 12 weeks exhibited poorer social competence scores than children who were breastfed for longer durations.

A similar study by Heikkilä and colleagues (2011) of over 10,000 mother-child pairs from the UK examined the association between breastfeeding duration and five dimensions of child behavior at age 5: emotional problems, conduct problems, hyperactivity, peer problems, and prosocial behavior deficits. The results revealed that children who breastfed for longer durations (e.g., 16 weeks or more) were reported as having fewer behavioral problems at age five across the five measured dimensions. In the case of full-term infants, these results were robust to the influence of various confounders, including socioeconomic status, maternal mental health, maternal age, maternal education, relationship status, smoking during pregnancy, and admission of offspring to a neonatal unit. A more recent study using a large sample of Irish children

(N=8568) yielded similar results (Reynolds et al., 2014). The findings indicated that,

48 independent of child obesity and a host of other confounding factors, initiating breastfeeding was associated with a 26% reduction in the risk of mental health and behavioral problems by age 9.

In addition to the evidence linking breastfeeding to offspring behavioral problems during childhood, a few studies have even detected evidence of a breastfeeding effect on behavior, personality, and mental health during adolescence and adulthood (Hayatbakhsh et al., 2012;

Merjonen et al., 2011; Oddy et al., 2010). For instance, Hayatbakhsh and colleagues (2012) detected a significant association between breastfeeding duration and adolescent psychopathology in their Australian sample (n = 4,502). To be precise, children who were still breastfeeding at 4 months of age were reported by their mothers to have significantly fewer mental health symptoms and problem behaviors at age 14 than those who failed to initiate breastfeeding or who terminated breastfeeding before 4 months of age. The association between breastfeeding duration and adolescent attention problems, social problems, and aggression was robust to various confounders, including substance use during pregnancy and maternal mental health. Another study of an Australian sample by Oddy and colleagues (2010) revealed that being breastfed for less than 6 months significantly predicted a) increased scores on both externalizing and total problem behavior domains through childhood and into adolescence and b) clinically significant levels of these behavioral problems through childhood and into adolescence, net of various maternal, family, and perinatal factors. There is also some initial evidence that a short duration of breastfeeding may contribute to a more hostile disposition during adulthood (see Merjonen et al., 2011).

Despite a general pattern of support for the link between breastfeeding duration and offspring externalizing behavior, the extent of supportive findings should not be overstated, as some studies have yielded conflicting evidence (Colen & Ramey, 2014; Kramer et al., 2008b;

49 Kwok et al., 2013; Lind et al., 2014; Shelton et al., 2011). A large, randomized trial of a breastfeeding promotion intervention implemented in the republic of Belarus, for instance, found no evidence that a shorter duration of breastfeeding contributed to greater total child difficulties, conduct problems, peer problems, hyperactivity, or diminished prosocial behavior at 6.5 years of age (Kramer et al., 2008b). Importantly, the study was unable to compare behavioral differences in children who initiated or failed to initiate breastfeeding (due to their similar presence in both the control group and the experimental group). Nonetheless, the results highlight the potential threat of selection bias, since the randomized trial yielded null results, whereas most observational studies find a significant effect. Other scholars have found evidence that the results of observational studies might be misspecified due to genetic confounding (Shelton et al.

2011). Using a prenatal cross-fostering design, Shelton and colleagues (2011) examined a sample of 870 families with a child aged 4-11 and found that, when the relationship between breastfeeding and offspring conduct problems was examined among genetically unrelated mother-child dyads, the effects were null. These results intimate the potential relevance of genetic confounding in the breastfeeding literature, or selection bias due to genetic factors, and the need for additional research using genetically informative designs.

Other scholars have noted that studies that find support for the association between breastfeeding duration and externalizing behavior may not be generalizable to non-western developed settings, such as China (Kwok et al., 2013). Nevertheless, the external validity of this body of findings seems to be less of a concern than their internal validity. To illustrate, cross- population generalizability of observational studies on the breastfeeding-externalizing link appears to be moderately good, with supportive studies taking place in Spain (Julvez et al.,

2007), the UK (Heikkilä et al., 2011), Ireland (Reynolds et al., 2014), Japan (Yorifuji et al.,

50 2014), Australia (Hayatbakhsh et al., 2012), Israel (Mimouni-Bloch et al., 2013), Turkey

(Sabuncuoglu et al., 2014), and America (Oddy et al., 2010). Overall, the breastfeeding- externalizing relationship appears to persist across several geographical locales. Nevertheless, a randomized control trial (Kramer et al., 2008b), a recent sibling study (Colen & Ramey, 2014), and a genetically informative study (Shelton et al., 2011) have cast doubt on whether the relationship between breastfeeding and conduct problems during kindergarten is causal or spurious. In light of the full body of literature, and the methodological shortcomings of the observational literature, an unequivocal endorsement of breastfeeding as a means of minimizing childhood behavioral problems is not currently possible.

4.2 Breastfeeding: Optimal Infant Nutrition or Facilitator of Attachment Security?

Breastfeeding is a unique, dyadic process in which various nutrients and growth factors are being transmitted from mother to child (Ballard & Morrow, 2013; Lönnerdal, 2013). At the same time, however, breastfeeding is more than the ingestion of nutrients; it is an inherently relational process that is characterized by skin-on-skin contact and cradling, and may also involve the mother gazing at and/or singing/speaking to her infant (see Liu et al., 2013). The particularly intimate and social nature of this form of feeding likely explains why the act of breastfeeding is frequently referred to as nursing. The mother-child intimacy that typically characterizes the breastfeeding experience might also be one of the key facilitators of attachment security. This possibility is further enhanced by research indicating that breastfeeding promotes greater activation in maternal brain regions responsible for empathy and bonding, which results in greater maternal sensitivity to offspring cues (Kim et al., 2011; see also Pearson, Lightman, &

Evans, 2011). Furthermore, breastfeeding might impact mother-infant attachment by increasing

51 maternal oxytocin levels, which would be expected to increase maternal positivity and nurturing

(Feldman et al., 2011; Nissen et al., 1996) and has been found to reduce maternal anxiety and depression during the postpartum period (Boutet, Vercueil, Schelstraete, Buffin, & Legros, 2006;

Stuebe, Grewen, & Meltzer-Brody, 2013). Consequently, it is possible that a closer mother- offspring attachment is facilitated by the act of breastfeeding, and that the closeness derived from the breastfeeding experience might ultimately impact the cognitive, socioemotional, and behavioral trajectory of breastfed children.

Interestingly, the findings of studies on the relationship between breastfeeding and the development of the mother-child bond are somewhat equivocal (see Brandt, Andrews, & Kvale,

1998; Britton et al., 2006; Else-Quest et al., 2003; Papp, 2013; Tharner et al., 2012). Ultimately, the research suggests that the relationship between breastfeeding, maternal sensitivity, and mother-infant attachment is complex. There is some evidence to suggest that mothers who breastfeed for longer durations show greater sensitivity to their 3-month-old infants than mothers who do not (Britton et al., 2006). Maternal sensitivity, in turn, seems to predict security of attachment at 12 months. Still, Britton and colleagues (2006) found that this relationship was explained by the prenatal intention to breastfeed, and not the act of breastfeeding itself, suggesting the potential relevance of maternal personality in her degree of maternal sensitivity

(see also Jones, 2013; Ekström & Nissen, 2006). A more recent study by Tharner and colleagues

(2012), however, detected positive associations between breastfeeding, maternal sensitivity, and attachment security, even though specific attachment classifications were not significantly associated with breastfeeding practices. Importantly, because breastfeeding enhances the maternal oxytocin response (see Feldman et al., 2011; Nissen et al., 1996; Strathearn et al.,

52 2012), which is associated with sensitivity to infant cues (Strathearn et al., 2009), the impact of breastfeeding on attachment security may be subtle and/or indirect.

Despite some supportive findings, a study by Else-Quest and colleagues (2003) found little evidence to buttress the claim that mother-infant dyads who engage in breastfeeding have a higher quality relationship than mother-infant dyads who engage in bottle-feeding. Similarly, more recent research has suggested that breastfeeding duration, while associated with maternal sensitivity, is not associated with the degree of positivity or negativity in mother-child interactions across early childhood (Papp, 2013). Furthermore, two recent reviews of the effects of breastfeeding on the mother-infant relationship concluded that, while there has been ample discussion of the theoretical mechanisms linking breastfeeding and the quality of the mother- child bond, empirical evidence supporting this association is generally lacking (see Jansen et al.,

2008; Schulze & Carlisle, 2010). The failure to find consistent associations may be due to divergent breastfeeding experiences of mothers. Qualitative research using intensive interviewing techniques has suggested that breastfeeding experiences can be quite divergent, with some women expressing a deep sense of intimacy and harmony with the infant while breastfeeding, and others reporting a disconnected, disrupted experience (Schmied & Lupton,

2001). However, in cases where breastfeeding is a rewarding social experience for both mother and child, breastfeeding appears to subtly enhance maternal mood and related prosocial behaviors during the feeding process (Bigelow et al., 2014; Ekström & Nissen, 2006; Feldman &

Eidelman, 2003), which are significantly associated with mother-infant attachment (Coyl,

Roggman, & Newland, 2002). Still, as noted previously, it may be that the relational benefits derived from breastfeeding do not persist past the earliest stages of the life course.

53 Ultimately, while the theoretical rationale behind the breastfeeding-attachment relationship is sound, the extent to which breastfeeding promotes a healthier mother-infant attachment than bottle-feeding remains relatively unclear. Still, if breastfeeding does indeed enhance maternal sensitivity and improve mother-child interactions as the child develops, then any link between breastfeeding and subsequent behavioral problems might be at least partly explained by the relationship quality of the mother-child dyad, and not the exposure to nutrients and growth factors in breast milk. Scholars who have explored the link between breastfeeding and offspring behavioral problems have frequently acknowledged this possibility without explicitly exploring it (Hayatbakhsh et al., 2012; Julvez et al., 2007; Mimouni-Bloch et al., 2013;

Reynolds et al., 2014; Yorifuji et al., 2014). For example, Hayatbakhsh and colleagues (2012) identify the child-mother relationship/interaction as a mechanism that might explain the link between breastfeeding duration and child behavioral problems, yet they do not overtly test this possibility.

Thus, while partial mediation may exist, research suggests that mother-infant attachment is unlikely to completely explain away the breastfeeding-externalizing relationship (see Heikkilä et al., 2011). Still, due to the lack of research that accounts for infant-mother attachment, definitive conclusions on the specific role of attachment security in the breastfeeding- externalizing relationship are difficult to draw, particularly since it is equally possible that optimal brain health/development is a key mechanism that explains the link between breastfeeding duration and behavioral outcomes (see Reynolds, 2001). For instance, there is evidence to suggest that the cognitive benefits derived from breastfeeding persist even when only expressed breast milk is provided to the infant (e.g., through nasogastric tube or a bottle) (see

Horwood, Darlow, & Mogridge, 2001; Lucas et al., 1992). Although currently unknown,

54 breastfeeding, independent of method of delivery, might confer similar benefits on the behavioral trajectories of children. Additional research is clearly needed to tease out the extent to which psychobiological (e.g., nutritional factors, hormones, neurotransmitters) and socioemotional

(e.g., low attachment security) elements explain the relationship between a short duration of breastfeeding and elevated conduct problems during childhood.

4.3 Attachment Security and Externalizing Behavior

Although it remains relatively unclear whether breastfeeding enhances the security of attachment of the infant to his/her caregiver, the association between an insecure attachment and childhood externalizing behavior is well-supported by the literature (Dubois-Comtois et al.,

2013; Groh et al., 2012; Kochanska & Kim, 2013; O’Connor et al., 2012). For example,

Kochanaska and Kim (2013) recently followed 101 children from age 15 months to 6.5 years.

The authors compared each child’s security of attachment to both mother and father at 15 months to the extent of their behavioral problems (e.g., bullies, lies, annoys, defies, etc.) at age 6.5. The results indicated that children who were insecurely attached to both their mother and father were rated by teachers as exhibiting a significantly greater degree of behavioral problems than children who were securely attached to both parents (or were securely attached to one). An additional study of 91 Portuguese children revealed that institutionalized children exhibited significantly lower security of attachment, and that attachment representations were in turn associated with a greater degree of behavioral problems, including aggression (Torres et al.,

2012). These results were robust to controls for parent’s education, age, and cognitive ability

(Torres et al., 2012). Ultimately, research using diverse samples suggests that externalizing

55 behavioral problems can persist through late childhood for children who exhibit an insecure attachment to their mother (see Groh et al., 2012; O’Connor et al., 2012).

The mechanisms by which low attachment security increases the risk of externalizing behavior are likely numerous. Attachment theory (Ainsworth, 1979; Bowlby 1969; 1982) proposes that a secure mother-infant attachment is characterized by a relationship in which the infant expects that their emotional needs will be met and/or managed by the caregiver, which provides the child with a functional working model that enables him/her to adapt to a host of challenges (see Andreassen & Fletcher, 2007; Calkins & Leerkes, 2004). Ultimately, attachment is reflected in the manner in which “children respond to sources of threat and challenge, and the extent to which children are able to draw on parental support and comfort as a means of coping”

(Fearon, Bakermans Kranenburg, Van IJzendoorn, Lapsley, & Roisman, 2010, p. 436). The theory also posits that‐ an internal working model of the expectation to rely on the caregiver for safety and comfort becomes generalized over time and ultimately impacts the child’s ability to function interpersonally across ever-widening social circles as they age. Thus, it is not surprising that scholars have repeatedly detected a link between low attachment security and externalizing behavior.

Ultimately, the theory also suggests that attachment is a key contributor to specific developmental domains, namely emotional regulation and morality/empathy, and the empirical literature is generally supportive of this claim (see Bost et al., 2014; Diener, Mangelsdorf,

McHale, & Frosch, 2002; Riva Crugnola et al., 2011; van der Mark et al., 2002). Each of these developmental domains, in turn, has been linked to the development of behavioral problems. For example, Calkins, Gill, Johnson, and Smith (1999) found evidence that 24-month-old toddlers with poor emotion regulation (e.g., high venting, high focal-object focus) were significantly

56 more likely to engage in aggressive conflict with peers in response to a distress-eliciting task.

Similar findings were garnered by Melnick & Hinshaw (2000) using a sample of boys with

ADHD. The authors detected less constructive patterns of emotional coping and regulation among high-aggressive subgroups relative to low-aggressive ADHD and nondiagnosed comparison subjects (Melnick & Hinshaw, 2000). Moral disengagement has also been linked to childhood behavioral problems. Gini, Pozzoli, and Hymel (2014), for example, recently studied a sample of nearly 18,000 8-18 year old boys and found a significant, positive association between moral disengagement and aggressive behavior among both children and youth. Taken together, these results suggest that the robust finding linking attachment to externalizing behavior is likely explained by multiple dimensions of cognitive, temperamental, and socioemotional development.

In sum, the link between attachment and externalizing behavior is quite robust, even though the link between breastfeeding and attachment is not entirely clear. If, however, a shorter duration of breastfeeding does interfere with attachment security, then breastfeeding measures may merely be serving as proxies for mother-infant attachment, which would suggest that breastfeeding practices may have little effect on externalizing behavior independent of the influence of attachment security on externalizing behavior. As mentioned previously, this possibility is occasionally acknowledged by breastfeeding scholars, but rarely subjected to empirical testing. Additional empirical attention to this possibility is necessary in order to determine the relevance of breastfeeding in the development of both mother-infant attachment and externalizing behavior.

57 4.4 Early Childhood Nutrition and Externalizing Behavior

As infants develop into toddlers, and toddlers into children, the specific content of their diet undergoes dramatic changes. Prior to the infant’s first birthday, solid foods are typically incorporated into the diet. The American Academy of Pediatrics (AAP) currently recommends that introduction of solid foods occur around 6 months of age (Kuo, Inkelas, Slusser,

Maidenberg, & Halfon, 2011). This transition is frequently made while the infant is still consuming a largely liquid diet of either breast milk or formula. Although some infants might be hesitant to be weaned, research indicates that repeated exposure to specific solid foods increases the infant’s acceptance and eventual consumption of that food (Sullivan & Birch, 1994).

Ultimately, the child is weaned off the breast and/or bottle completely and proceeds to consume a diet that consists entirely of solid foods. By early childhood, the child’s diet is likely to consist of foods regularly consumed by the average adult (e.g., fruits, breads, meats, vegetables, etc.), albeit in smaller quantities.

Although the size of the brain increases tremendously during infancy and toddlerhood

(Hanson et al., 2013), important neurological changes continue to occur during early childhood.

Children are progressing in their emotional and cognitive sophistication during this time due to normative brain development. This is influenced by processes in which a variety of neuronal networks become more organized (Boersma et al., 2013). In their review of the literature, Brown and Jernigan (2012) recently noted that brain development during the preschool years is incredibly dynamic, exhibiting changes in morphology, white matter tracts, and connective tissue. To be precise, they assert that, during this stage of the life course, “the brain shows some of its largest annualized changes in both its anatomical and physiological characteristics” (Brown

& Jernigan, 2012, p. 326). Importantly, research has revealed that various events and

58 environments can alter the developmental trajectory of the brain during early childhood, perhaps partly through epigenetic changes (Fox, Levitt, & Nelson, 2010). Various researchers have found evidence that early childhood constitutes a “critical” or “sensitive” period for the formation of certain neuronal connections (Kolb & Gibb, 2011; Rapoport & Gogtay, 2007) and that, overall, “there is more striking evidence for plastic brain changes in childhood” relative to later life stages (Rapoport & Gogtay, 2007, p. 182).

Specifically, it has been suggested that future modification of the circuits that mature earliest (i.e., those that process lower level information) may be difficult if not achieved during the appropriate window, as relevant networks will become “perceptually biased” (Fox et al.,

2010; see also Kolb & Gibb, 2011). The implication is that younger, less mature neuronal networks are, for better or worse, more plastic (i.e., more sensitive to environmental influence) than older, more trained networks (Greenough, Black, & Wallace, 1987; Kolb & Gibb, 2011).

Whatever the exact neurological mechanisms, it is clear that various neurocognitive skills are still developing and maturing during early childhood, and that such skills can be enhanced or hampered by various environmental stimuli, including nutritional factors (Kolb & Gibb, 2011).

A large body of research has linked various nutritional factors during early childhood to features of brain development (Golding, Emmett, Iles-Caven, Steer, & Lingam, 2014; Kitsao-

Wekulo, Holding, Taylor, Kvalsvig, & Connolly; 2013; Nyaradi et al, 2013). For example, a recent review by Nyaradi and colleagues (2013) indicates that various micronutrients, including omega-3 fatty acids, iron, iodine, zinc, and vitamin B12, play an essential role in facilitating the neuropsychological development of children. Furthermore, inadequate exposure to such nutrients from pregnancy though childhood can result in significant impairments in cognition. A study by Kitsao-Wekulo et al. (2013) revealed that poor nutritional status during childhood

59 significantly contributes to the development of deficits in language skills, motor skills, and executive functioning. Notably, each of these neuropsychological outcomes has been linked to an increased risk of low self-control (Beaver, DeLisi, Vaughn, Wright, & Boutwell, 2008;

Jackson & Beaver, 2013; Ratchford & Beaver, 2009; Scheres et al., 2004) and associated behavioral problems such as misconduct and delinquency (Cauffman, Steinberg, & Piquero,

2005; Espy et al., 2011; Jackson & Beaver, 2013; Piquero & White, 2003). It is reasonable, therefore, to suspect that nutritional deficits during early childhood might increase the risk of aggressive and externalizing behavior, at least in part by impacting neuropsychological health.

4.4.1 Malnutrition and Externalizing Behavior

To date, a relatively small number of studies have examined whether malnourished children are at greater risk of exhibiting conduct problems years into the future (Galler et al.,

2011; Liu et al., 2004; Liu & Raine, 2006). So far, studies have typically detected a significant association between malnutrition during early childhood and subsequent behavioral problems.

To illustrate, Liu et al. (2004) followed a sample of 1,795 children from the republic of Mauritius from the age of 3 until the age of 17. Signs of malnutrition in a subset of the children were detected at 3 years of age, including hair dyspigmentation, hair loss, and angular stomatitis. In comparison to controls, the malnourished children (N = 353) exhibited more aggressive and hyperactive behaviors at age 8, more externalizing behaviors of age 11, and signs of conduct disorder at age 17, independent of psychosocial adversity. The results also indicated that neurocognitive deficits mediated the link between early childhood malnutrition and subsequent misconduct. A review of the literature by Liu and Raine (2006) suggests that inadequate nutrition does indeed play a role in the development of behavioral problems in children.

60 However, they also note the need for a greater number of randomized control trials to better elucidate the utility of prevention and intervention programs.

More recent studies examining the link between malnutrition and externalizing behavior appear to confirm the results of Liu and colleagues (2004). For example, Galler and colleagues

(2011) found that early childhood malnutrition (defined as moderate-to-severe protein-energy malnutrition) was predictive of both deficits in executive functioning and higher parent-reported aggression towards peers at ages 9-15. Results were robust to controls for age, sex, maternal depression, and household standard of living. A follow-up study of Barbadian youth suggests that self-reported conduct problems are higher among individuals with a history of early childhood malnutrition, and that IQ partially mediates this relationship (Galler et al., 2012). A very recent study by Liu and colleagues (2014) extended the prior literature by exploring the extent to which deficiencies in specific micronutrients (e.g., zinc and iron) predict behavioral problems in preschoolers. Venous blood samples of 1314 Chinese children were collected to determine micronutrients levels, and the Child Behavior Checklist (CBCL) was employed to detect the extent of behavioral problems during preschool. The results suggest that, independent of key sociodemographic factors, low zinc levels, as well as low zinc and iron levels in combination, were associated with a greater degree of total behavioral problems. Nevertheless, iron levels did not independently predict behavioral problems in children.

4.4.2 Low Diet Quality and Externalizing Behavior

In addition to acute and micronutrient malnutrition, it is possible that the actual quality of diet, or differences in the frequency with which certain foods are consumed, might also influence the development of behavioral problems in children. A number of studies have investigated whether children with poorer eating habits are at an increased risk of conduct problems,

61 including ADHD symptomatology (Benton, 2007; Benton, 2008; Howard et al., 2011; Oddy et al., 2009; Oh et al., 2013; Park et al., 2012; Woo et al., 2014). Overall, the results indicate that a western dietary pattern is particularly conducive to the development of externalizing behavior

(Oh et al., 2013; Park et al., 2012; Woo et al., 2014). For example, a principal components analysis conducted by Woo and colleagues (2014) yielded evidence of four distinct dietary patterns among their Korean sample of elementary school children: the “snack” pattern, the

“traditional” pattern, the “traditional-healthy” pattern, and the “seaweed-egg” pattern.

Additional analyses revealed that the traditional-healthy pattern of eating, characterized by a diet low in fat and high in fatty acids and minerals, lowered the odds of developing ADHD, whereas the snack pattern, characterized by high consumption of sweets, snacks, and breads, increased the odds of developing ADHD. Another study of a Korean sample of preschoolers found that, among girls, a pattern of eating that included a high level of sweets was associated with poorer behavior and social skills (Oh et al., 2013).

Similar results have been obtained using samples of different ages and nationalities. For example, research has indicated that Australian youths who consume more fast food, red meat, french fries, soft drinks, and sugars (i.e., a western diet) are 121% more likely to be diagnosed with ADHD, after adjusting for numerous confounders from pregnancy to 14 years of age

(Howard et al., 2011). Similarly, a study by Oddy and colleagues (2009) used a 212-item food frequency questionnaire to assess dietary patterns in an Australian sample of adolescents. They also found that a western dietary pattern (high intake of fast food, sugars, and red meat) was associated with greater externalizing behavioral problems, including delinquency and aggression.

Conversely, the results indicated that more frequent consumption of fresh fruit and leafy green vegetables corresponded to improvements in behavior scores.

62 There have also been a handful of randomized control trials examining the benefits of comprehensive micronutrient supplementation in curbing externalizing behavior (Sinn & Bryan,

2007; Sinn, 2008). A review of this research by Sinn and Bryan (2007) suggests that behavioral problems such as hyperactivity, inattention, and impulsivity can be mitigated when diets are supplemented with polyunsaturated fatty acids (PUFAs) (see also Sinn, 2008). Namely, omega-

3 and omega-6 PUFAs, including docosahexaenoic acid (DHA) and arachidonic acid (AA), appear to impact behavior due to their central role in the development and functioning of the central nervous system and the brain (Schuchardt, Huss, Stauss-Grabo, & Hahn, 2010). Still, other randomized control studies have found that, in additions to PUFAs, various other micronutrients, from vitamin A and iodine to iron and zinc, can play a role in the behavioral profiles of both adults and children (Schoenthaler & Bier, 2000; for a review, see Benton, 2008).

For example, whether children (Kaplan, Fisher, Crawford, Field, & Kolb, 2004; Schoenthaler &

Bier, 2000), adolescents (Schoenthaler, 1983, 1985; Schoenthaler et al., 1997) or adults (Gesch,

Hammond, Hampson, Eves, & Crowder, 2002; Zaalberg, Nijman, Bulten, Stroosma, & van der

Staak, 2010) are examined, providing nutrient-dense diets through supplementation appears to significantly reduce various forms of antisocial behavior, including fighting, vandalism, endangering others, and other aggressive behaviors.

In sum, acute malnutrition, micronutrient malnutrition, and poor dietary patterns all appear to heighten the risk of various behavioral problems (e.g., externalizing behavior) during childhood and even into later life stages. Conversely, there is some evidence to suggest that mimicking a well-balanced diet through the use of supplementation can reduce the frequency and/or severity of conduct problems in children, adolescents, and adults.

63 4.5 Is Early Childhood Nutrition Associated with Attachment Security?

As noted previously, the link between nutritional factors during infancy and attachment security is tenuous at best, despite a fairly reasonable rationale behind the expected association.

Another related, yet distinct question is whether attachment security during the first years of life is predictive of the subsequent eating behaviors of children. Although the amount of research conducted on the topic is somewhat limited, studies to date indicate that children who are not securely attached to their caregiver are at risk of poor dietary practices (see Bost et al., 2014;

Bozorgi et al., 2014; Faber & Dubé, 2015; Simmons et al., 1995; for an excellent summary, see

Lu et al., 2013). To illustrate, a very recent study by Bost and colleagues (2014) examined 497 primary caregivers and their 2.5-3.5 year-old children. The results revealed that fewer parental attachment behaviors corresponded to a negative emotional regulation style in offspring. Both an insecure attachment and negative emotional regulation, moreover, were found to place children at risk of unhealthful food consumption (e.g., low fruit and vegetable consumption; high fast food, salty snacks, and sweets). Other studies have examined attachment-based interventions as a method of altering eating behaviors in overweight children. For instance,

Bozorgi and colleagues (2014) recently conducted an experimental trial including 32 obese elementary students and found that a ten-week attachment-based intervention significantly improved the eating behaviors of the experimental group, relative to the control group. Some studies have even indicated that the development of a secure attachment during the first few years of life can significantly influence long-term preventive health behaviors (e.g., vegetable, fruit, and fast food intake) (see Huntsinger & Luecken, 2004).

Lu and colleagues (2013) argue that primary caregivers and food are the two most central objects of life during the first few years of development. Both of these objects have the potential

64 to provide emotional reinforcement to the child. To the extent that a secure attachment between the parent and the child is established early in life, it is expected that the attachment style will become internalized, providing a protective mechanism that lays the foundation for healthy eating patterns across the life course (Lu et al., 2013). In short, the authors propose that the internalization of the secure attachment style obviates the need to repeatedly turn to unhealthy foods for their emotional-reinforcing properties, which results in a general pattern of healthy eating that is driven through a biological reinforcement process. Their results are consistent with their hypothesis: parent-child relationships characterized by less trust and more anxiety are associated with a pattern of unhealthful food consumption, regardless of the age of the offspring

(Lu et al., 2013). These results are consistent with the argument that an insecure attachment increases the risk of poor eating patterns by interfering with the ability of individuals to adequately regulate their emotions (Anderson et al., 2012; also see Bost et al., 2014, for an example). This process likely also explains why low attachment security has repeatedly been linked to childhood obesity (see Anderson et al., 2012; Anderson & Whitaker, 2011; Trombini et al., 2003).

In light of the research linking childhood diet quality to attachment security (Bost et al.,

2014; Bozorgi et al., 2014; Faber & Dubé, 2015; Goossens et al., 2011), and attachment security to externalizing behavior (Dubois-Comtois et al., 2013; Groh et al., 2012; Kochanska & Kim,

2013; O’Connor, et al., 2012), it is possible that the associations detected between diet quality and externalizing behavior (see Oh et al., 2013; Park et al., 2012; Wiles et al., 2007; Woo et al.,

2014) may be explained by antecedent levels of parent-child attachment. Scholars exploring the impact of diet on childhood behavior have yet to account for attachment security as a potential confounding mechanism, opting instead to account only for parent demographics such as age,

65 education, and occupation (see Woo et al., 2014, for an example). For example, while Park and colleagues (2012) detected a significant link between a western dietary pattern of processed fat and meat intake and ADHD-related behavioral problems, the authors did not account for the role of attachment security. The authors do, however, acknowledge that children’s poor eating habits may be related to child temperament and/or inappropriate parenting (see also Oh et al., 2013).

Still, it appears that researchers examining the link between nutrition and behavior have yet to consider the potential importance of attachment security in the relationship between diet and externalizing behavior.

4.6 The Role of Genetic Factors in the Development of Externalizing Behavior

In addition to the effects of various environmental and socioemotional factors (e.g., breastfeeding, diet, attachment style), research has revealed that genes contribute to a host of antisocial outcomes (Arseneault et al., 2003; Baker et al., 2007; Barnes et al., 2013; Beaver et al.,

2007; Deater-Deckard & Plomin, 1999, Mota et al., 2013; Zai et al., 2012), including childhood externalizing behavior (Arseneault et al., 2003; Baker et al., 2007; Barnes et al., 2013; Brendgen et al., 2005; Saudino, Ronald, & Plomin, 2005). A number of meta-analyses have indicated that, in general, approximately 50% of the variance in antisocial phenotypes is attributable to genes

(see Ferguson, 2010; Mason & Frick, 1994; Miles & Carey, 1997; Rhee & Waldman, 2002).

When examining aggression and related externalizing behaviors among child samples, heritability estimates tend to be a little higher. For example, Barnes and colleagues (2013) found that between 78 and 82% of the variance in externalizing behavior during preschool and kindergarten was genetic in origin, whereas a study by Baker and colleagues (2007) detected even higher levels of heritability (.96). Still, some research suggests that different

66 subdimensions of externalizing behavior might have differing degrees of heritability. For example, a study by Polderman, Posthuma, De Sonneville, Verhulst, and Boomsma (2006) examined teacher ratings of various child problem behaviors and found that attention/hyperactivity problems were 63% heritable, whereas aggression and rule breaking were approximately 45% heritable.

Interestingly, research that employs broad, general measures of childhood antisocial behavior from multiple sources (e.g., parents, teachers, child care providers) detects the strongest influence of genes on externalizing behavior (for an example, see Baker et al., 2007). For example, Arseneault et al. (2003) and Baker et al. (2007) used reports from multiple informants and garnered heritability estimates ranging from .82 to .96, whereas Polderman and colleagues

(2006), who employed only teacher reports, detected notably lower, although still significant, levels of heritability (.45-.63). This finding suggests that genes may be even more influential than previous meta-analyses have suggested, since the few studies using highly comprehensive measures of childhood antisocial behavior tend to both a) glean the full spectrum of the child’s behavior across settings and b) detect the highest levels of heritability (see Arseneault et al.,

2003; Baker et al., 2007). Furthermore, studies incorporating physical aggression, as opposed to only social aggression, into their measures of antisocial behavior are more likely to detect significant heritability estimates (see Brendgen et al., 2005, for an example).

Apart from studies that estimate latent measures of genetic risk for externalizing behavior, a number of scholars have also linked specific genetic polymorphisms to various antisocial behaviors (Beaver et al., 2007; Mota et al., 2013; Zai et al., 2012), including childhood externalizing behavior (Beitchman et al., 2006; Zai et al., 2012). For the most part, the genes most consistently linked to persistent aggression and externalizing problems during childhood

67 are implicated in neurotransmission. For instance, research has recently revealed that several dopaminergic genes, including DAT1, DRD2, and DRD4, are implicated in childhood aggression (Zai et al., 2012). Furthermore, research has suggested that DRD2 and DRD4 may work together to influence conduct problems (Mota et al., 2013). 5-HTTLPR (or the serotonin transporter polymorphism) has also been linked to extreme, persistent aggression during childhood. Specifically, individuals who possess “low expressing” genotypic variants (e.g., S/S, or two short alleles) are significantly more likely to exhibit severe behavioral problems

(Beitchman et al., 2006). Some research has indicated that other genetic loci associated with social behavior, emotions, and human bonding might also influence childhood aggression, including two SNPs on the oxytocin receptor gene OXTR (i.e., rs6770632 and rs1042778) (see

Malik, Zai, Abu, Nowrouzi, & Beitchman, 2012) and one SNP on the vasopressin receptor

(AVPR) 1B (i.e., rs35369693) (Zai et al., 2012). Ultimately, additional research that examines a wider array of candidate genes is needed. Nevertheless, both the behavioral and molecular genetic research to date examining childhood externalizing behavior is supportive of a link between genetic factors and childhood behavioral problems.

4.7 The Role of Perinatal Factors in the Development of Externalizing Behavior

Although genetic factors clearly play a role in the development of externalizing behavior, a wealth of research has also revealed that disadvantageous circumstances surrounding the time of birth, including various perinatal insults (e.g., birth complications, low birth weight, etc.), significantly increase the risk of childhood behavioral problems (Beaver & Wright, 2005;

Breslau et al., 1996; Elgen, Sommerfelt, & Markestad, 2002; Hack et al., 2009; Hultman et al.

2007; Linnet et al., 2006; Mick et al., 2002; Nigg & Breslau, 2007). Many of these studies

68 reveal that children weighing less than 2500 grams at birth (in some cases, less than 1500 or

1000 grams) are at an increased risk of ADHD and related symptomatology (e.g., hyperactivity, impulsivity, and inattention) (Hack et al., 2009; Hultman et al. 2007; Jackson & Beaver, 2015;

Linnet et al., 2006; Nigg & Breslau, 2007). For example, a recent study by Hack and colleagues

(2009) found that children born at extremely low birth weights exhibit the highest severity scores for inattentive, hyperactive, and combined types of attention-deficit hyperactivity disorder at eight years of age. The link between low birth weight and ADHD symptomatology appears to be robust to genetic controls (Hultman et al. 2007) as well as a host of familial controls, including maternal education, socioeconomic status, prenatal smoking, and family history of psychiatric disorders (see Linnet et al., 2006; Nigg & Breslau, 2007).

Although this line of research typically examines the link between perinatal risk and

ADHD symptomatology, a number of studies have also specifically linked myriad perinatal risk factors to aggression and violence during both childhood and adolescence (Arseneault,

Tremblay, Boulerice, & Saucier, 2002; Buschgens et al., 2009; Liu et al., 2009; Raine, Brennan,

& Mednick, 1994). For example, Arseneault and colleagues (2002) found that subjects whose births involved induced labor, umbilical cord prolapse, and/or maternal symptoms of preeclampsia were significantly more likely to exhibit violent behavior at ages 6 and 17. Other complications, including fetal distress, breech birth, and premature rupture of membrane, have also been linked to aggression during late childhood (see Liu et al., 2009). Furthermore, research has also revealed significant associations between obstetrical complications and violent/criminal behavior during adulthood (Kandel & Mednick, 1991; Tibbetts & Piquero, 1999), particularly in the presence of maternal rejection (see Raine et al., 1994; Raine, Brennan. & Mednick, 1997; see also Beck & Shaw, 2005).

69 Importantly, Raine and colleagues (1997) found the effects of birth complications to emerge only in the case of early-onset violence (as opposed to late-onset violence), which highlights the particular relevance of perinatal factors for the development of physical aggression during the early stages of the life course. Liu (2011) recently argued that certain complications at birth, such as hypoxia or anoxia, likely increase the risk of aggressive and externalizing behaviors by damaging hippocampal formation and interconnected aspects of the limbic system, areas of the brain that are associated with low self-control and aggression (Liu, 2004; Liu &

Wuerker, 2005). Liu’s hypothesis is further corroborated by the findings of Beaver and Wright

(2005), which suggest that severe oxygen deprivation at birth significantly reduces childhood levels of self-control. While more research distinguishing the effects of different perinatal risk factors is clearly warranted, there is a general consensus in the research to date that events that occur around the time of birth have the potential to influence subsequent behavior by shaping early brain development (Liu, 2011; Liu et al., 2009).

4.8 A Biosocial Process: Differential Vulnerability to Poor Nutrition

Despite the literature examining the role of both infant and early childhood nutrition in the development of cognitive and behavioral outcomes, relatively few studies have considered the possibility that different children may have distinct cognitive and behavioral responses to poor nutrition, depending on the presence of certain preexisting risk factors (Anderson et al.,

1999; Caspi et al., 2007; Morales et al.,, 2011; Quigley et al., 2012). In short, different children may respond differently to the risks incurred by poor nutrition. For example, possessing a high degree of genetic risk might make children especially sensitive to a shorter duration of breastfeeding or a diet low in essential nutrients during childhood. Ultimately, it is possible that

70 genetic, perinatal, socioemotional, and nutritional risk factors work in conjunction to heighten the risk of behavioral problems during childhood.

The expectation that these factors might interact to predict behavioral problems is rooted in a large body of literature that buttresses the role of biosocial interactions in the development of antisocial behaviors (Beaver, 2008; Caspi et al., 2002; Moffitt, 2005; Reif et al., 2007), including child externalizing behavior (Bakermans Kranenburg & Van IJzendoorn, 2006; Petkovsek,

Boutwell, Beaver, & Barnes, 2014; Propper,‐ Willoughby, Halpern, Carbone, & Cox, 2007). For example, Petkovsek and colleagues (2014) recently found that the influence of prenatal smoking on offspring externalizing behavior is heightened in the presence of genetic risk. Furthermore, research has revealed that the development of children who are exposed to various perinatal risk factors (e.g., pre-term birth, low birth weight, and/or birth complications) may be especially impacted by several environmental risk factors, including a short duration of breastfeeding

(Quigley et al., 2012), maternal rejection (Raine, et al., 1997), low SES (Tibbetts & Piquero,

1999), disadvantageous family structure (Tibbetts & Piquero, 1999), and family adversity (Beck

& Shaw, 2005).

To the extent that similar biosocial interactions are detected between nutritional factors and other genetic, perinatal, and socioemotional moderators, it could lead to improved identification of infants and children who are at the greatest risk of developing an early onset of antisocial behavior, and thereby enhance prevention/treatment efforts at the earliest stages of life.

Although research on the possible role of breastfeeding-gene interactions in the development of externalizing behavior problems is virtually non-existent (except see Groen-Blokhuis et al.,

2013), a handful of studies to date have explored whether the effect of breastfeeding duration on cognition is moderated by the genetic polymorphism rs174575 in the FADS2 gene (Caspi et al.,

71 2007; Morales et al.,, 2011; Rizzi et al., 2013). In 2007, Caspi and colleagues hypothesized that allelic variation on FADS2, a gene that is implicated in the regular of fatty acid pathways in the brain, would moderate the effects of breastfeeding on IQ, as fatty acids were posited as the main mechanism by which breastfeeding improves child cognition. Specifically, breastfed children who possess one or more C alleles on rs174575 in the FADS2 gene scored significantly higher on IQ tests (6.4 points on average) relative to breastfed children with no C alleles (i.e., G allele homozygotes). The results suggest that the gene may play a role in maximizing the degree of fatty acids in breast milk that are absorbed by the brain. Follow-up studies have demonstrated that genetic variation in the FADS2 gene does indeed influence the expression of fatty acids in the brain (Rizzi et al., 2013). A recent study by Morales and colleagues (2011) replicated the evidence of a breastfeeding-gene interaction, finding that the effect of breastfeeding on cognition is moderated by genetic variation in fatty acid enzymes that are encoded by FADS2.

The interaction, however, has not been consistently detected by research to date (Groen-

Blokhuis et al., 2013; Steer et al., 2010). Furthermore, only one study has examined whether the breastfeeding-FADS2 interaction has any bearing on behavioral outcomes. Groen-Blokhuis and colleagues (2013) found no evidence of either a direct effect of FADS2 on behavioral problems or an interactive effect with breastfeeding duration, despite finding small, beneficial effects of breastfeeding on cognition and behavior. Thus, it remains relatively unclear whether genetic variation in FADS2 leads to differential cognitive and behavioral outcomes for breastfed children. Importantly, no other polymorphisms or indicators of genetic risk have been considered as moderators of the breastfeeding-externalizing relationship (except see Jackson &

Beaver, in press, for a related study). It is also possible that indicators of genetic risk might moderate the relationship between early childhood nutrition and externalizing behavior. It is

72 widely recognized that a poor quality diet during childhood might represent a more serious risk factor for obesity and other bodily ailments (e.g., diabetes), depending on genotype (Cornelis &

Hu, 2012; Garver et al., 2013). Nevertheless, the possibility that genetic risk and poor nutrition might interact to predict behavioral problems has, to my knowledge, never been subjected to empirical testing. Importantly, much of the research examining gene-nutrient interactions explores their role in the cognitive (Dauncey, 2012; Mattson, 2003; Whalley et al., 2008) and physical health (Karnehed et al., 2006; Mathers & Hesketh, 2007; Ordovas, 2006) of older subjects, failing to consider whether similar interactions are relevant to the behavioral patterns of young children.

In terms of the potential moderating role of perinatal risk factors in the relationship between breastfeeding and child externalizing behavior, a number of studies have suggested that low birth weight infants have the most to gain from a longer duration of breastfeeding, and that breastfeeding may help them to “catch up” developmentally to their normal birth weight peers

(see Horwood et al., 2001; Lucas et al., 1992; Vohr et al., 2007). Furthermore, research has indicated that breastfeeding has a stronger effect on the cognitive development of low birth weight children relative to normal birth weight children (Anderson et al., 1999; Quigley et al.,

2012). Research to date has not explicitly tested whether a short duration of breastfeeding is associated with higher levels of externalizing behavior during kindergarten among children who were born low birth weight. Still, it is reasonable to suggest that, if children who are born low birth weight are more apt to benefit from breast milk when it comes to their cognitive functioning, a similar process may be occurring in the case of behavioral outcomes as well.

Conversely, in light of the interconnectedness of cognition and behavior during childhood

(Saarinen, Fontell, Vuontela, Carlson, & Aronen, 2014; Schoemaker et al., 2013), it is reasonable

73 to suggest that the influence of birth weight on behavioral problems might be heightened when the duration of breastfeeding is short (or when breastfeeding is never initiated). Again, scholars have yet to test this possibility. Still, to the extent that adequate nutrition is more essential to the functioning of the brain of low birth weight children, it stands to reason that nutrition during infancy may be more influential in the behavior of low birth weight children than normal birth weight children.

Finally, while various early childhood processes may moderate the relationship between nutritional factors and externalizing behavior, one that has yet to be thoroughly explored is attachment security. The mother-child bond, however, appears to enhance the protective effect of breastfeeding in the case of internalizing behaviors (see Liu and colleagues, 2013).

Specifically, Liu and colleagues (2013) studied a cohort of 1267 Chinese children at 6 years of age and found that children who experienced both exclusive breastfeeding and active mother- infant bonding exhibited the lowest risk of internalizing behavioral problems (e.g., anxiety, depression, and withdrawal). It appears reasonable, therefore, to suggest that nutritional factors and attachment might interact in a similar way to predict other kinds of problem behaviors. For example, breastfeeding might be especially useful in protecting against the externalizing behavior of offspring when it is accompanied by mother-child processes that facilitate the internalization of a secure attachment style. Conversely, failing to initiate breastfeeding, or breastfeeding for a short duration, may place the offspring at risk of externalizing behavior problems if the mother-infant pair are simultaneously unable to develop a secure attachment though other means, independent of breastfeeding (e.g., quality time/interaction, parental sensitivity). Of course, these hypotheses have yet to be tested, but they follow the same logic as

74 the Liu and colleagues (2013) study, which detected a biosocial process involving both bonding and nutritive elements.

To my knowledge, no research has extended the logic outlined above to examine the interactive effects of nutrition and attachment on behavioral outcomes during subsequent life stages, such as early childhood. Nevertheless, it is reasonable to suggest that the externalizing behavior of an insecurely-attached child might be exacerbated even further by a consistently poor diet. To be precise, since a low diet quality impedes brain health (see Gómez-Pinilla, 2008;

Molteni et al., 2002), and both a poor diet and low attachment security appear to interfere with emotional regulation (Anderson et al., 2012; Bost et al., 2014; Diener et al., 2002; Riva Crugnola et al., 2011), it would not be unexpected for the cognitive and socioemotional risk factors to accumulate and result in a particularly poor behavioral repertoire (see, for example, the biopsychosocial model posited by Dodge & Pettit, 2003; see also Teisl & Cicchetti, 2008).

In sum, there is ample precedent for the role of biosocial interactions in the development of antisocial behavior (Beaver, 2008; Caspi et al., 2002; Moffitt, 2005; Reif et al., 2007).

Additionally, a handful of studies to date have suggested that the influence of certain nutritional factors (e.g., breastfeeding during infancy) on cognitive outcomes may not be uniform across all levels of genetic and perinatal risk (see Anderson et al., 1999; Caspi et al., 2007; Morales et al.,,

2011; Anderson et al., 1999; Quigley et al., 2012). Even so, much remains unknown about the potential for genetic, perinatal, and socioemotional risk factors to moderate the relationship between nutritional factors and externalizing behavior problems during kindergarten.

75 4.9 Nutritional Interplay and the Formation of Externalizing Behavior

Typically, researchers who explore the association between nutritional factors and various child outcomes do not examine the interplay between nutritional factors at distinct life stages. To illustrate, while decades of research has linked breastfeeding to a host of child outcomes, from intelligence (Lucas et al., 1992) to obesity (Dewey, 2003), only a handful of recent studies have explored whether breastfeeding exposure and/or duration predicts subsequent eating habits and related health behaviors (see Abraham et al., 2012; Grieger et al., 2011; Perrine et al., 2014; Scott et al., 2012). Scholars who have examined this possibility generally find an association between nutritional factors during infancy (e.g., breastfeeding duration) and nutritional factors during later life stages. For instance, a recent study of large cohort of Scottish children found that children who were breastfed were more likely than those who were not to have a positive eating pattern at around 2 years of age (Abraham et al., 2012). Breastfed children, for example, were more likely to eat a wide variety of fruits and vegetables, and consume sweets and soft drinks less frequently. Although this relationship was attenuated after adjusting for SES, it remained significant, with the odds of a healthy eating pattern being 48% higher for breastfed children relative to non-breastfed children (see also Perrine et al., 2014, for similar results). Similarly, the odds of being overweight at age 4 were 26% lower for breastfed children compared to non-breastfed children.

Other recent studies intimate that the connection between breastfeeding and subsequent eating practices may extend beyond the first few years of life. A recently study by Grieger and colleagues (2011) examine 2,287 2-8 year-old children and found that those who had been breastfed were significantly more likely to exhibit a healthy pattern of eating, which was characterized by regular consumption of lean meats, vegetables, fruits, and wholegrain, and rare

76 consumption of take-away foods and carbonated beverages. Scott and colleagues (2012) argue that such findings are likely explained by the range of flavors transferred in breast milk, which provide “early exposure to different tastes and positively shape children’s food preferences and food variety” (p. 1464). Whatever the exact mechanism, there is some initial evidence that the process of breastfeeding can subtly influence offspring dietary patterns. In light of the relationship between poor diet and poor behavior, it is reasonable to suggest that breastfeeding may exert an indirect effect on childhood behavioral problems through its impact on eating behaviors. Still, empirical tests of the interrelationships between breastfeeding, childhood diet, and externalizing behavior are absent from the literature.

Moreover, scholars have not considered the possibility that breastfeeding and childhood eating patterns may interact to predict externalizing behavior. Never initiating breastfeeding, for example, has the potential to subtly impact brain development (Ballard & Morrow, 2013;

Lönnerdal, 2013), including the development of key brain structures, such as the superior and inferior parietal lobes (see Kafouri et al., 2012). Any potential differences in brain structure and functioning between breastfed and non-breastfed children may differentially prime them to be either more or less sensitive to diet quality in the life stages that follow. For example, a healthy diet during early childhood may more or less compensate for subtle neurological disadvantages that a child may have incurred as a result of not being breastfed. Similarly, little to no exposure to breastfeeding may have little impact on the behavioral outcomes of children who consume a wide variety of healthy foods during early childhood. Conversely, non-breastfed children may be especially vulnerable to behavioral problems to the extent that they also engage in poor eating habits during childhood. Although these types of interactive relationships are merely speculative, they are rooted in scholarly research, which has revealed that a) child development

77 is predicated on what was accomplished during prior life stages (for an example, see Rowe,

Raudenbush, & Goldin Meadow, 2012) and b) children’s cognitive functioning can render them differentially vulnerable‐ to risk and/or protective factors in their current environment (Ghods,

Kreissl, Brandstetter, Fuiko, & Widhalm, 2011; Jaekel, Pluess, Belsky, & Wolke, 2014; Luu,

Vohr, Allan, Schneider, & Ment, 2011; Wade et al., 2011). These developmental process have been detected, for example, in the case of low birth weight infants, who, under advantageous circumstances (e.g., breastfeeding), are likely to “catch-up” both physically and cognitively with the normal birth weight counterparts (see, for example, Ghods et al., 2011; Luu et al., 2011).

Under disadvantageous circumstances (e.g., maternal insensitivity), however, the development of the low birth weight infant can remain stunted (see Jaekel et al., 2014). In a similar fashion, it is possible that children with little to no exposure to breastfeeding may catch up developmentally with children who had a greater degree of exposure to breastfeeding if they consume a highly nutritious diet during early childhood. Conversely, if exposure to nutritional risk factors accumulates across infancy and early childhood, children may be especially prone to develop behavioral problems.

Ultimately, development is an ongoing process, and the behavioral patterns that result are likely the culmination of the unique interaction of risk and protective factors, both environmental and genetic, that came before. The objective of the current study is to distinguish the unique and interactive contribution of nutritional, genetic, socioemotional and perinatal factors in the development of externalizing behavior by employing a genetically informative sibling design.

78 CHAPTER 5

METHODS

Prior chapters have provided a statement of the problem in the current literature (Chapter

1), an overview of the empirical and theoretical developments concerning the early childhood origins of antisocial behavior (Chapter 2), a more thorough treatment of the key limitations in the current nutrition literature and the contribution of this dissertation (Chapter 3), and a comprehensive, multidisciplinary review of the literature examining both the relationship between nutrition and externalizing behavior and the biosocial/developmental mechanisms that are relevant to the present research questions (Chapter 4). I will now describe the data that are employed in the current study, the operationalization of key constructs, and the analytical techniques that are used to address the research questions at hand.

5.1 Data

The current study uses data from the Early Childhood Longitudinal Study, Birth Cohort

(ECLS-B). The ECLS-B examines a large, nationally-representative sample of children born in the United States in 2001. Using a stratified sampling approach, ECLS-B researchers sampled birth certificates registered with the National Center for Health Statistics in the year 2001, which covers approximately 99% of U.S. births that occur in a given year. Children were deemed ineligible if a) they died before the age of 9 months b) they were adopted before the age of 9 months or c) their mothers were younger than 15 at the time of birth.

Five waves of data have been collected to date, spanning several years of development

(i.e., from age 9 months until the kindergarten school year). Data were collected from multiple sources, including parents, independent raters, fathers, day care providers, and school teachers.

79 Additionally, direct assessments of children cognitive skills as well as birth certificate data were obtained. Approximately 10,600 children participated in the study at the first wave of data collection.

Interviews at wave 1 were conducted between the fall of 2001 and the fall of 2002, when the children were, on average, about 9 months of age (although they ranged from about 6 to 14 months of age). The primary caregiver (which was typically the mother) was asked a range of questions regarding their own mental health, accomplishments, and well-being, as well as questions regarding family life and the focal child’s temperament, development, and behavior.

During wave 1, a subsample of fathers was also interviewed about the well-being of their child(ren). Direct motor and mental assessments of the children, as well as direct observations of parent-children interactions, were also obtained at this wave. Importantly, questions regarding the circumstances surrounding the birth of the child, including early feeding questions, were assessed at this wave. Furthermore, data concerning perinatal complications and birth weight were garnered from each focal child’s birth certificate.

The second wave of data collection occurred between the fall of 2003 and the fall of

2004, when the children were approximately 2 years old. Mothers were asked a range of questions about their relationship status, educational attainment, involvement with their child, parenting techniques, antisocial history, socioeconomic well-being and mental health. They were also asked about the degree to which their child’s cognition, temperament, and behavior were developing normally (i.e., developmental milestones). Because a number of subjects were involved in one or more child care arrangements at this age, data pertaining to child care arrangements were collected for a subsample of subjects through phone interviews with day care providers as well as direct observational assessments of child care conditions.

80 During wave 3 of data collection, which took place between the fall of 2005 and the fall of 2006, many survey items were modified in order to reflect the enhanced autonomy and sophistication of the focal children. For example, parents were asked several questions pertaining to their child’s academic preparedness, social aptitude, learning capacities, and behavioral control. To be precise, parents were asked about the child’s propensity toward aggression, low impulse control, attention problems, temper tantrums, and destructiveness.

Again, data pertaining to child care arrangements were collected through phone interviews with center care providers as well as direct observational assessments of center care conditions.

Finally, in the fall of 2007, the fourth wave of data were collected from the remaining sample of children who had entered kindergarten, which constituted approximately 75% of the sample. Subjects who had not yet entered kindergarten by 2007, or who had entered kindergarten in 2007 but were repeating it, were assessed on the same measures during the fifth wave of data collection, which took place the following school year (i.e., the fall of 2008).

Importantly, during waves 4 and 5 of data collection, teachers were asked to report on their qualifications, teaching style, and classroom setting, as well as the traits and behaviors of the focal children. Specifically, questions regarding the learning, temperament, behavior, and peer relationships of focal children were asked of teachers at waves 4 and 5. Similar questions regarding the behavior of focal children were also asked of parents at these waves.

5.1.1 Analytical Sample

The ECLS-B is especially well-suited to the current study due to its inclusion of a large sample of approximately 1,600 twins. Twins were oversampled in the ECLS-B study, which enables researchers to conduct genetically informative analyses of the data. Only same-sex twin pairs were evaluated to determine their zygosity (i.e., whether they were monozygotic or

81 dizygotic twins), since opposite-sex twins are always dizygotic. Both parents and independent raters were asked to evaluate the similarity of the twins on several indicators (i.e., hair texture, eyes color, complexion, ear lobes, etc.). Furthermore, the blood type and Rh factors of each twin were ascertained through parental report. The process used to determine zygosity has been widely used, as it is highly reliable and valid (Cohen, Dibble, Grawe, & Pollin, 1975; Goldsmith,

1991). After eliminating opposite-sex twins and twins with undetermined zygosity, the sample employed in the current study consisted of nearly 1000 twins (N = 976), 238 monozygotic and

738 dizygotic. Importantly, final sample sizes of specific models will vary contingent on the measures included in each model and their degree of missingness.

5.2 Measures: Breastfeeding Analyses

5.2.1 Externalizing Behavior (W4/5)

At waves 4 and 5 of data collection, both parents and kindergarten teachers were asked a number of questions concerning the behavioral, cognitive, and emotional abilities of focal children. Measures were taken from the Preschool and Kindergarten Behavior Scales – Second

Edition (PKBS-2) (Merrell, 2003). Most of the questions included in the PKBS-2 were asked of both parents and teachers, with a few exceptions. Specifically, while parents were asked all 24 questions included in the assessment, teachers were only asked 22 of the questions. Nonetheless, the breadth of topics covered is comparable in the parent and teacher reports.

Of the 46 questions, seven parent-rated questions and six teacher-rated questions were identified as indicators of externalizing behavior. Principal components analyses and factor analyses of these sets of items reveal that one factor solution was detectable for each set of items

(see also Barnes et al., 2013). Furthermore, the inclusion of additional items from the PKBS-2

82 does not improve the observed alpha level. Parents were asked seven questions regarding their child’s externalizing behaviors. In particular, they were asked about how often the child got angry, acted impulsively, was unable to sit still, and engaged in physically aggressive acts (e.g., hit, kick, or punch) during the 3 months prior to the interview. Parents were also asked how frequently the child threw tantrums, destroyed things, and annoyed other children during the 3 months prior to the interview. Response options ranged from 1 (never) to 5 (very often).

Teachers were asked very similar questions at waves 4 and 5. Specifically, teachers were asked to report on the extent to which the child acts without thinking, engages in physical aggression, is overly active, disrupts other children and/or the class, annoys/bothers other children, and has temper tantrums. Response options for these six items also ranged from 1

(never) to 5 (very often).

Because 25% of the sample was assessed at wave 5, but not at wave 4, a combined measure of waves 4 and 5 was created to examine kindergarten externalizing problems, regardless of the wave at which the child entered kindergarten. Specifically, children who did not have valid data at wave 4 were included in the analysis by utilizing their data from wave 5

(i.e., the wave they entered kindergarten). However, for children who had already entered kindergarten at wave 4, the wave 4 items were used.1 Both parent-rated and teacher-rated items were summed together and averaged to create a scale of externalizing behavior at wave 4/5.

Importantly, the internal reliability of the items was high (alpha = .86). The scale was created so that higher scores on the scale are indicative of a greater manifestation of externalizing behavioral problems across school and family settings.

1 About 5% of the same-sex twin sample repeated kindergarten. However, because this did not significantly impact externalizing behavior in our sample, it was not included as a covariate in the analyses. Only a handful of twin pairs differed in their timing of entry into kindergarten. Although this predicted significant differences in externalizing behavior (b = .49, p = .02), such differences did not change the results of the present study in any substantive way.

83 5.2.2 Breastfeeding Measures

Short Duration of Breastfeeding. At waves 1 and 2 of data collection, a series of filter questions were asked concerning the breastfeeding of each focal child in the home. To be precise, the primary caregiver (approximately 99% of whom were mothers in the same-sex twin sample) was asked “Did (you/the child’s mother) ever breastfeed (the child)”? Response options included yes (1) or no (2). Respondents who answered yes were then asked if they (or the mother) were (was) still breastfeeding the child at the time of the interview, with response options being yes (1) or no (2). Respondents who answered no (2) to this item were then asked

“For how many months did (you/the child’s mother) breastfeed (the child)?” These same items, or slight variations of these items, were asked again at wave 2, in order to capture the variation in the duration of breastfeeding for those children who were still breastfeeding at the first wave of data collection, but had stopped by the second wave. Data from waves 1 and 2 were used to construct a variable that measured the duration of breastfeeding (in months) for each child.

Subjects whose mothers reported (at wave 1 or wave 2) that they had never breastfed them were assigned a value of 0 on this variable. Children who had terminated breastfeeding by the first wave of data collection were assigned the value corresponding to the number of months of breastfeeding reported by the caregiver/mother. The same process was used to assign values for children who were still breastfeeding at the first wave of data collection, but had terminated breastfeeding by the second wave of data collection.

Importantly, for the same-sex twin sample, children’s ages at wave 2 ranged from 22 to

32 months, although 98% of children fell between the ages of 23 and 28 months. Children who were still breastfeeding at wave 2, however, were all at least 23 months old. As a result, focal children whose mothers/caregivers reported that they were still breastfeeding at wave 2 were

84 assigned a value of 23. However, it is important to note that being assigned a value of 23 on this variable signifies, in reality, 23 months or more of breastfeeding, as the data do not provide a means of determining the precise duration of breastfeeding for this group of children.

Additionally, children who were no longer breastfeeding at wave 2, but had discontinued breastfeeding at the age of 23 months or more, were also assigned a value of 23 on this variable.

Ultimately, because children who had not yet aged beyond 23 months by wave 2 were never given the opportunity to report breastfeeding past 23 months, children whose mothers/caregivers reported that breastfeeding was terminated at an age beyond 23 months were necessarily assigned a value of 23, along with children who were still breastfeeding at wave 2 (all of whom were at least 23 months old). It should be noted that, of the full same-sex twin sample, only 17 children were reported as being breastfed for 23 months or more (less than 2% of the sample).

Therefore, this slight truncation of the continuous breastfeeding measure likely has little to no bearing on the validity of the results. On the contrary, this dose-response measure of breastfeeding is an improvement over most measures used in prior studies, at least in terms of nuance and specificity of duration (see Evenhouse & Reily, 2005; Hayatbakhsh et al., 2012;

Heikkilä et al., 2011; Reynolds et al., 2014).

In order to create a continuous measure of short duration of breastfeeding, the number of months (ranging from 0-23) during which the child was breastfed was reverse coded, so that individuals who were never breastfed were assigned a value of 23, and individuals who breastfed for 23 months or more were assigned a value of 0. Each score that fell between 0 and 23 was also reverse coded by subtracting it from 23. Thus, higher scores on this measure are indicative of a shorter duration of breastfeeding. The clear utility of this continuous measure is found in its

85 ability to test for dose-response effects of breastfeeding (i.e., whether incremental, month-to- month changes in breastfeeding influence externalizing behavior).

Breastfed Less than 6 Months. In addition to a continuous measure of breastfeeding, I also created a dichotomous measure of breastfeeding that measured whether or not each focal child had been breastfed for at least 6 months. Prior research has suggested that breastfeeding may have unique benefits on health, cognition and even behavior once it crosses the 6-month threshold (Mimuoni-Bloch et al., 2013; Niegel, Ystrom, Hagtvet, & Vollrath, 2008; Oddy et al.,

2010; Quinn et al., 2001; Wigg et al., 1998). Furthermore, the World Health Organization has recommended that women exclusively breastfeed their offspring for at least the first 6 months of life (See Kramer & Kakuma, 2012). The recommendation, however, may be in need of reappraisal, in light of recent evidence (including randomized control trials) suggesting that the influence of breastfeeding for 6 months has been overstated (see Fewtrell, Wilson, Booth, &

Lucas, 2011; Kramer et al., 2008b). In order to determine whether the WHO recommendations are justified, subjects who were never breastfed, or were breastfed for less than 6 months, were assigned a value of 1, whereas subjects who breastfed for 6 months or more were assigned a value of 0.

Not Exclusively Breastfed. The final breastfeeding measure constructed for the current study is a binary measure that taps whether each child was exclusively breastfed for 6 months or more. This was determined by simultaneously employing information concerning a) the child’s breastfeeding history, b) the child’s formula feeding history and c) the child’s cow’s milk and solid food history. Children were assigned a value of 1 on this variable if they met any of the following conditions: 1) were breasted for less than 6 months, 2) were fed formula prior to 6 months of age, 3) were fed cow’s milk prior to 6 months of age, and/or 4) were fed solid foods

86 prior to 6 months of age. Subjects that met none of the four above conditions were assigned a value of 0. The inclusion of this measure of breastfeeding exclusivity in the current study is largely motivated by the World Health Organization’s emphasis on exclusively breastfeeding for

6 months in order for the offspring to derive the most benefits out of breastfeeding (See Kramer

& Kakuma, 2012). Some studies suggest that behavior can improve as a result of long-term, exclusively breastfeeding (see Yorifuji et al., 2014, for an example). Others suggest that exclusive breastfeeding can impact various physical health outcomes (see Duijts, Jaddoe,

Hofman, & Moll, 2010; Kramer et al., 2003; Ladomenou, Moschandreas, Kafatos, Tselentis, &

Galanakis, 2010). Still, some researchers have challenged the claim that exclusive breastfeeding, relative to a nonexclusive approach, has any unique benefits for health or cognition, let alone behavior (see Fewtrell et al., 2011; Kramer et al., 2007). The issue of whether the exclusivity of breastfeeding is as important as the duration of breastfeeding remains inconclusive.

Consequently, I decided to compare duration, threshold (< 6 months), and exclusivity effects in order to hopefully detect which process, if any, is most predictive of particular patterns of behavior during kindergarten.2

5.2.3 Demographic Control Variables

Age. A continuous measure (in months) of each respondent’s age during their kindergarten year.

Race. A binary measure of race in which nonwhites are coded as 1 and whites are coded as 0.

Sex. A binary measure of sex in which males are coded as 1 and females are coded as 0.

Low Household Income. At the first wave of data collection, parents were asked to indicate their approximate household income per year. Response options ranged from 0 (less

2 For additional details on the exact items used to create each of the breastfeeding measures, see APPENDIX A.

87 than $5000) to 12 (more than $200,000), with each response option indicating a range of income values. This item was reverse coded so that higher scores on the variable reflected a lower household income.

5.2.4 Additional Control Variables

Age of Mother. The age of the mother was also ascertained at the first wave of data collection. Mothers were asked to report their current age in years.

Postpartum Depression. At wave 1, when their infants were approximately 9 months old, mothers were asked 12 questions designed to tap the extent to which they experienced symptoms of depression. Questions were asked about the extent to which mothers felt unfocused, lonely, sad, bothered, and unmotivated. Mothers were also asked about their eating and sleeping habits.

Items were summed and averaged to create a postpartum depression scale (alpha = .87). The scale was constructed in such a way that higher scores on the scale corresponded to a greater degree of postpartum depressive symptomatology.

Low Maternal Education. At wave 1, mothers were asked to report their highest level of education attained. Response options ranged from 1 (8th grade or below) to 9 (doctoral or professional degree). Items were reverse coded so that higher scores reflected lower educational attainment.

Female-Headed Household. Features of the family structure were also ascertained at wave 1. Children who were living in a home in which the mother was present but the father was absent were assigned a value of 1, whereas all other children were assigned a value of 0.

Low Parental Involvement. Mothers also reported on their involvement with their child at the first wave of data collection. Specifically, mothers were asked about the frequency with which they read to their child, told stories to their child, and sang to their child (alpha = .66).

88 Responses to these items were coded so that higher scores were indicative of less parental involvement. The reverse-coded items were summed and averaged to construct a low parental involvement scale.

5.2.5 Moderating Variables

Low Attachment Security.3 At the second wave of data collection, when the focal children were approximately 2 years of age, the development of attachment relationships between parents and offspring was assessed using the Toddler Attachment Sort-45 Item (TAS-

45), which is a modified version of the Attachment Q-Sort (AQS) (Andreassen & Fletcher, 2007;

Waters & Deane, 1985). The TAS-45 is an assessment tool designed to detect the extent to which children are able to use their parent (typically the mother) as a secure base from which they feel free to explore the world, yet confident to return to in times of distress (see Bowlby

1969 and Bowlby, 1982, for more on attachment theory). The TAS-45 is intended to tap the degree to which the child’s internalized working model of the self, and the self in relation to others, has become secure within the first 2 years of life.

The process of measuring low attachment security was completed by the interviewer during the home visit that occurred at the second wave of data collection. Assisted by a computer, interviewers, who underwent extensive training prior to data collection, were charged with the task of sorting 45 cards, first into two piles (i.e., “applies” or “not applies”), but ultimately into four piles (from, “almost always applies” to “rarely or hardly ever applies”).

These observations were recorded over the course of approximately 2 hours, with sorting lasting roughly 10 minutes. Listed on the cards were child behaviors, and corresponding parent-child interactions, that are reflective of the degree of attachment security (e.g., “Child seeks and enjoys

3 Importantly, low attachment security is also employed as a control, outcome, and mediating variable in some analyses.

89 being hugged by mother”).4 These behaviors can be summarized using the following nine dimensions, or “hot spots”: is comfortably cuddly, is cooperative, enjoys company, is independent, seeks attention, is upset by separation, avoids others/not sociable, is demanding/angry, is moody/unsure/unusual. In general, children who are less securely attached to their caregivers engage in fewer warm, cooperative, and/or sociable behaviors and more clingy/anxious, demanding, angry, and/or moody behaviors (see Bronte-Tinkew, Scott, &

Horowitz, 2009; Roisman & Frayley, 2008, for additional details). The pattern of traits is illustrated in figures B.1 and B.2 (located in Appendix B).

In order to tap the quality of the target child’s attachment relationship with the caregiver,

ECLS-B researchers created a continuous attachment security measure that was derived from the results of the sorting task. This continuous measure is provided in the ECLS-B data, and is indicative of the degree to which the child is prototypically secure.5 This measure was obtained by calculating correlations between the resulting score on the attachment sort (from all nine dimensions) and a criterion sort representing a hypothetical child with ideal levels of security

(see Andreassen & Fletcher, 2007, for more details). Scores on this measure ranged from -1 to 1:

1 representing the prototypically secure child, and -1 representing the prototypical insecure child.

Thus, children exhibiting behaviors most consistent with secure attachment score closer to 1, and children exhibiting few, if any, behaviors consistent with secure attachment score close to -1.

For the current study, however, this item was reverse coded, so that higher scores reflect a lower degree of attachment security. Notably, a number of prior studies using the ECLS-B have employed this same comprehensive measure as an indicator of attachment security (see

Anderson & Whitaker, 2011; Bronte-Tinkew, Scott, & Horowitz, 2009; Roisman & Frayley,

4 For a complete list of items, please see APPENDIX B. 5 For additional details on the attachment measure, see APPENDIX B.

90 2008) and a comprehensive meta-analysis has buttressed the validity of the measure (see Van

Ijzendoorn, Vereijken, Bakermans-Kranenburg, & Riksen-Walraven, 2004).

Low Birth Weight. The ECLS-B also contains information concerning the immediate circumstances surrounding the birth of each child. For the current study, I employ an indicator of low birth weight obtained from birth certificate data as an indicator of perinatal risk. Focal children who were born weighing less than 2500 grams (i.e., 5 lbs., 8 oz.) were assigned a value of 1 and all other children were assigned a value of a 0.

Genetic Risk. A measure of genetic risk for externalizing behavior was also created for the current study. Importantly, molecular markers of genetic risk were not available in the

ECLS-B, precluding tests of interactions between nutritional measures and specific high-risk polymorphisms. Nevertheless, the inclusion of twin pairs in the ECLS-B permits the construction of a latent measure of genetic risk. I followed the lead of prior research (see

Beaver, Barnes, May, & Schwartz, 2011; Jaffee et al., 2005; Petkovsek et al., 2014 ; van Lier et al., 2007) and created a measure of genetic risk that is a function of a) each twin’s zygosity and b) their cotwin’s score on externalizing behavior. In order to do this, I first randomly designated one twin within each twin pair to be the target twin, and the other twin within each twin pair to be the cotwin. Next, the cotwin’s score on the externalizing behavior scale (W4/5) was dichotomized at the 90th percentile, in order to identify cotwins who exhibited particularly high levels of externalizing behavior. To be precise, cotwins with an externalizing behavior score at or above the 90th percentile were assigned a value of 1, whereas cotwins with a score below the

90th percentile were assigned a value of 0.

Following the creation of a binary variable that identifies the cotwins with the most severe behavioral problems, genetic risk can be modeled as a function of twin zygosity. In this

91 sense, a target twin who is monozygotic and has a cotwin with relatively severe behavioral problems (i.e., at or above the 90th percentile) will score the highest on the measure of genetic risk (3). Target twins who are dizygotic and have a cotwin with relatively severe behavioral problems will score the next highest on the measure of genetic risk (2). Target twins are assigned a value of 1 on the genetic risk variable when they are dizygotic, but their cotwin was assigned a value of 0 on the dichotomous measure of externalizing behavior. Finally, monozygotic target twins whose cotwin was assigned a 0 on the dichotomous measure of externalizing behavior are given the lowest score on the genetic risk variable (0). Thus, the constructed genetic risk variable ranges from 0 to 3, with higher value reflecting a greater degree of genetic risk.

Importantly, this method of determining genetic risk rests on the logic that the only reason MZ twins, as a group, should be more similar to each other than DZ twins on a given phenotype is because MZ twins share, on average, twice as much of their distinguishing DNA.

To be precise, this logic is employed in the creation of the genetic risk scale, since it was constructed to give higher scores to target twins with both a) a higher degree of genetic similarity to their cotwin and b) a cotwin with particularly high levels of externalizing behavior. The logic that underpins the creation of this measure rests on an assumption known as the Equal

Environments Assumption (EEA), which suggests that “the environmental factors that are etiologically relevant to a given phenotype are no more likely to be shared by MZ twin pairs than

DZ twins pairs” (LoParo & Waldman, 2014, p. 606).

A recent meta-analysis of research on the EEA by Barnes and colleagues (2014) revealed that the vast majority of empirical examinations of the EEA support it (see also Cronk et al.,

2002; Kendler, Neale, Kessler, Heath, & Eaves, 1993). In cases where there is evidence of bias,

92 the average degree of bias is minimal (upward bias of .05). Even so, Barnes and colleagues

(2014) conducted a number of simulations which revealed that, on the whole, any upward bias incurred as a function of violating the EEA is counterbalanced by the downward bias due to assortative mating practices (see also a comprehensive evaluation of the EEA by Felson, 2014).

Another recent study by LoParo & Waldman (2014) specifically examined the validity of the

EEA as it pertains to twin research on childhood externalizing behavior. The authors found that a number of environmental factors and externalizing symptoms of MZ twins were more similar than those of DZ twins. Even so, the strength of cross-twin correlations in externalizing symptoms did not vary by the level of rearing environment similarity. In short, the environments that were shared to a greater degree by MZ twins did not explain the greater concordance in the externalizing behaviors of MZ twins. These findings yield support for the EEA, since the environmental influences that were more similar among MZ twins than DZ twins were not etiologically relevant to the development of externalizing behavior.

5.3 Measures: Low Diet Quality Analyses

5.3.1 Externalizing Behavior (W4/5)

The same measure of externalizing behavior that is used as an outcome variable in the breastfeeding analyses is also used as an outcome variable in the low diet quality analyses. For further details on the construction of this measure, see “5.2.1 Externalizing Behavior (W4/5).”

5.3.2 Early Childhood Diet Measures

Low Vegetable Consumption. At wave 3 of data collection, when subjects were approximately 4-5 years of age, caregivers were asked about the dietary patterns of their children. Specifically, several components of the children’s diet were tapped in a series of

93 questions in order to determine each child’s eating habits. In the current study, the first component of the diet examined was vegetable consumption. Parents were asked how many times during the past seven day their child ate vegetables, excluding french fries and other fried potatoes. Parents were also told that, in addition to raw vegetables, cooked vegetables, such as those that are served in a stir fry, soup, or stew, should also be counted. Response options included not at all in the past 7 days (7), 1-3 times during the past 7 days (5), 4 to 6 times during the past 7 days (6), 1 time per day (1), 2 times per day (2), 3 times per day (3), and 4 or more times per day (4). In order to tap low vegetable consumption as a measure of nutritional risk, this item was coded so that higher scores were given to children whose parents reported that they ate vegetables less frequently. To be precise, participants whose parents responded not at all in the past 7 days were assigned a value of 6 (the highest possible value) and participants whose parents responded four or more times a day were assigned a value of 0 (the lowest possible value). Thus, possible scores on this item ranged from 0 to 6.6

Low Fruit Consumption. In addition to the question about vegetable consumption, another question regarding the frequency of fruit consumption of the focal children was also asked of the parents at wave 3. Specifically, parents were asked, “During the past 7 days, how many times did your child eat fresh fruit, such as apples, bananas, oranges, berries or other fruit such as applesauce, canned peaches, canned fruit cocktail, frozen berries, or dried fruit?”

Respondents were explicitly told that fruit juice should not be included when determining their answer to the question. Response options ranged from never in the past 7 days to four or more times a day. In the same manner as the vegetable consumption variable, this item was coded so that higher scores were given to children whose parents reported that they ate fruit less frequently. To be precise, participants whose parents responded not at all in the past 7 days

6 For additional details on the exact items pertaining to preschool diet, see APPENDIX A

94 were assigned a value of 6 (the highest possible value) and participants whose parents responded four or more times a day were assigned a value of 0 (the lowest possible value).

High Fast Food Consumption. Parents were also asked how often their child ate fast food during the past 7 days. In particular, parents were asked, “During the past 7 days, how many times did your child eat a meal or snack from a fast food restaurant with no wait service such as McDonald’s, Pizza Hut, Burger King, Kentucky Fried Chicken, Taco Bell, Wendy’s and so on?” Importantly, respondents were explicitly told that they should include fast food consumption, regardless of where the food was actually eaten (e.g., in the fast food establishment, in the car, at home). Response options ranged from never in the past 7 days to four or more times a day. This item was coded so that higher scores were given to children whose parents reported that they ate fast food more frequently. To be precise, participants whose parents responded not at all in the past 7 days were assigned a value of 0 (the lowest possible value) and participants whose parents responded four or more times a day were assigned a value of 6 (the highest possible value).

High Sweets Consumption. Parents were also asked how frequently their child ate sweet/sugary foods during the past 7 days. Specifically, parents were asked, “During the past 7 days, how many times did your child eat candy (including Fruit Roll-Ups and similar items), ice cream, cookies, cakes, brownies, or other sweets?” Response options ranged from never in the past 7 days to four or more times a day. This item was coded so that higher scores were given to children whose parents reported that they ate sweets more frequently. To illustrate, participants whose parents responded not at all in the past 7 days were assigned a value of 0 (the lowest possible value) and participants whose parents responded four or more times a day were assigned a value of 6 (the highest possible value).

95 High Salty Snack Consumption. Parents were also asked how frequently their child ate salty snacks during the past 7 days. Specifically, parents were asked, “During the past 7 days, how many times did your child eat potato chips, corn chips such as Fritos or Doritos, Cheetos, pretzels, popcorn, crackers or other salty snack foods?” Response options ranged from never in the past 7 days to four or more times a day. This item was coded so that higher scores were given to children whose parents reported that they ate salty snack foods more frequently. For instance, participants whose parents responded not at all in the past 7 days were assigned a value of 0 (the lowest possible value) and participants whose parents responded four or more times a day were assigned a value of 6 (the highest possible value).

High Soda Consumption. The final early childhood nutrition question pertains to the types of beverages the child drinks on a regular basis. In particular, parents were asked how frequently their child drank soda pop (such as Coke, Pepsi, etc.) or other sugary drinks that are not 100% fruit juice (e.g., Sunny Delight, Hi-C, Kool-Aid). Response options ranged from never in the past 7 days to four or more times a day. This item was coded so that higher scores were given to children whose parents reported that they drank soda and other sugary drinks more frequently. To be precise, participants whose parents responded not at all in the past 7 days were assigned a value of 0 (the lowest possible value) and participants whose parents responded four or more times a day were assigned a value of 6 (the highest possible value).

Low Diet Quality. In an effort to approximate the overall eating patterns of focal children in the study, I created a composite measure of low diet quality by summing the previous six items together. As each item has a possible range of 0-6, and there are six items, scores on the composite item can range anywhere from 0 to 36. However, in the same-sex twin subsample, the lowest observed score was 2 and the highest observed score was 29. Subjects who scored higher

96 on this composite measure are those who have the least healthy eating habits across the six diet domains tapped in the current study (i.e., low vegetable consumption, low fruit consumption, high fast food consumption, high sweets consumption, high salty snack consumption, high soda consumption), whereas those who scored lower have healthier overall eating habits.

5.3.3 Demographic Control Variables

Age. A continuous measure (in months) of each respondent’s age during their kindergarten year.

Race. A binary measure of race in which nonwhites are coded as 1 and whites are coded as 0.

Sex. A binary measure of sex in which males are coded as 1 and females are coded as 0.

Low Household Income (W3). At the third wave of data collection, parents were asked to indicate their approximate household income per year. Response options ranged from 0 (less than $5000) to 12 (more than $200,000), with each response option indicating a range of income values. This item was reverse coded so that higher scores on the variable reflected a lower household income.

5.3.4 Additional Control Variables

Low Maternal Education (W3). At wave 3, mothers were asked to report their highest level of education attained. Response options ranged from 1 (8th grade or below) to 9 (doctoral or professional degree). Items were reverse coded so that higher scores reflected lower educational attainment.

Maternal Depression (W3). At wave 3, when their children were approximately 4 to 5 years old, mothers were asked 12 questions designed to tap the extent to which they experienced

97 symptoms of depression. Questions were asked about the extent to which mothers felt unfocused, lonely, sad, bothered, and unmotivated. Mothers were also asked about their eating and sleeping habits. Items were summed and averaged to create a maternal depression scale

(alpha = .90). The scale was constructed in such a way that higher scores on the scale corresponded to a greater degree of depressive symptomatology reported by the mother.

Female-Headed Household (W3). Features of the family structure were also ascertained at wave 3. Children who were living in a home in which the mother was present but the father was absent were assigned a value of 1, whereas all other children were assigned a value of 0.

Parental Withdrawal (W3). Parents were also asked about the extent to which they experienced feelings of disengagement and frustration in their role as a parent at wave 3. Five items in particular were summed and averaged to create a parental withdrawal scale (alpha =

.78). Parents were asked about the degree to which they felt trapped in their role as a parent, felt tired, felt that parenting is difficult, felt that parenting involves more work than pleasure, and felt unwilling to sacrifice their own interests to meet their child’s needs. Items were coded so that individuals who reported a greater degree of withdrawal-related feelings received higher scores on the scale.

Corporal Punishment (W3). During the third wave of data collection, parents reported the extent to which they used physical force to punish their child during the week prior to the interview. During this wave, interviewers reminded parents that “Sometimes kids mind pretty well and sometimes they don’t” and then asked them the following question: “About how many times, if any, have you spanked {Child/Twin} in the past week for not minding?” Parents were told that answers must range from 0 to 90. Answers were recoded so that parents who engaged in any spanking during the course of the week prior to the interview were assigned a value of 1,

98 whereas parents who did not engage in any spanking during the course of the week prior to the interview were assigned a value of 0.

Infrequent Family Meals (W3). During the third wave of data collection, parents were asked to provide information concerning the frequency with which they and their children participate in family meals together. Specifically, parents were asked how many days at least some of the family ate their evening meal together in a typical week. Response options ranged from 0 days a week to 7 days a week. This item was reverse coded so that higher scores reflected a lower frequency of family meals in the average week.

No Family Food Rules (W3). A measure of a lack of structure and routine concerning the consumption of food in the household was derived from a question asked of parents at the third wave of data collection. To be precise, parents were asked, “In your house, are there rules or routines about what kinds of food {CHILD} {and {TWINS}} eat(s)?” Response options were simply yes (1) or no (0). The item was reverse-coded so that households with no family food rules or routines were assigned a 1, and those with rules/routines were assigned a 0.

Externalizing Behavior (W3). In order to control for concurrent externalizing behavior in the models of early childhood nutrition at wave 3, a measure of parent-rated externalizing behavior was included (alpha = .74). This measure exactly mirrors the parent-rated portion of the wave 4/5 measure of externalizing behavior. Specifically, parents were asked seven questions regarding the externalizing behavior of the focal children at wave 3, when the children were 4-5 years of age. The questions tapped the extent to which children were aggressive, impulsive, angry, and annoying to other children. Questions about the destruction of property, tantrums, and over-activity were also asked. Response options range from 1 (never) to 5 (very often). Higher scores on the scale indicate a greater degree of externalizing problems at wave 3.

99 5.3.5 Moderating Variables

Low Attachment Security.7 The same measure of low attachment security that is employed as a moderating variable in the breastfeeding analyses is also used as a moderating variable in the low diet quality analyses. For further information, see “5.2.5 Moderating

Variables.”

Genetic Risk. The same measure of genetic risk that is used as a moderating variable in the breastfeeding analyses is also employed as a moderating variable in the low diet quality analyses. For additional details concerning the construction of this measure, see “5.2.5

Moderating Variables.”

5.4 Plan of Analysis

The analyses for the current study are divided into three distinct sections designated by form of nutrition. The results of the analyses pertaining to breastfeeding are presented first, followed by the results pertaining to preschool diet quality, followed by the results pertaining to the interplay of nutritional factors across infancy and early childhood. Each of these sections proceeds as follows. First, descriptive statistics of all of the variables used in any of the analyses in that section are presented. Second, bivariate correlations between all of the variables used in any of the analyses in that section are presented. Third, the results of several difference of means t tests are presented to illustrate the extent to which parents, children, and households differ on covariates according to their nutritional patterns. Fourth, the results of several genetically informative tests of the direct and indirect links between nutritional factors, low attachment security, and externalizing behavior are presented. Finally, genetically informed product-term

7 Importantly, low attachment security is also employed as a control, outcome, and mediating variable in some analyses.

100 analyses are conducted to examine whether the interactions between nutritional factors and genetic, perinatal, and socioemotional factors are predictive of externalizing behavior during kindergarten.

5.4.1 Univariate and Bivariate Analyses

In order to glean the overall pattern of participants’ scores on the variables of interest, the means, standard deviations, and ranges are reported for all variables included in any of the analyses. This initial step in the analytical process provides an overview of the general patterns of the data, including the value around which subjects tend to cluster for each of the variables, as well as the extent of dispersion. In addition to the univariate analyses, two distinct types of bivariate analyses are also conducted. First, bivariate correlations between variables in the both the breastfeeding analyses and the low diet quality analyses are calculated in an effort to detect significant, zero-order associations between the variables of interest.

Second, difference of means t tests are employed to determine the degree to which features of households and their residents (e.g., mothers, children) are significantly different across distinct nutritional profiles. These tests separate the sample into two groups on the basis of a nutrition threshold (e.g., breastfed less than 6 months) to determine whether household, maternal, and child covariates significantly differ across this threshold (at the bivariate level).

These analyses are specifically tailored to address the first research question of each section of the analyses (i.e., research questions 1, 5, and 9). The logic underpinning the use of t tests is to establish a clear need for a within-family, sibling design in each section of the analyses. To the extent that nutritional factors are correlated with household/maternal covariates (i.e., shared environments), within-family analyses will enhance the researcher’s ability to rule out confounding due to these shared environmental factors. Research to date, for example, has

101 indicated that children with the poorest eating patterns are more likely to reside in economically and socially disadvantageous contexts (Casey, Szeto, Lensing, Bogle, & Weber, 2001; Hanson,

Neumark-Sztainer, Eisenberg, Story, & Wall, 2005). It is crucial, therefore, to determine whether similar patterns exist in the current data. If they do, it would justify the use of a sibling design that more effectively rules out alternative explanations of the potential associations between dietary factors during infancy and preschool and externalizing behavior during kindergarten. Using a sibling design, as a follow up to significant t test results, will hopefully improve the internal validity of this body of literature and minimize the likelihood of spurious results.

5.4.2 Defries-Fulker Analysis

There are two genetically informative techniques used in the present study. The first of these techniques is known as Defries-Fulker (DF) analysis. DF analysis is a regression-based method that permits the estimation of the relative effects of genetic factors, shared environmental factors, and nonshared environmental factors. These estimates are obtained by using samples of sibling pairs who differ in their degree of genetic similarity (e.g., MZ and same-sex DZ twins).

DF analysis decomposes the variance in the outcome variable into the proportions explained by genetic and environmental factors, while also allowing for the estimation of regression coefficients for specified nonshared environments (i.e., environments that are not shared by siblings within a kinship pair). I chose to utilize DF analysis for the current study for two main reasons. First, DF analysis is an effective way to control for residual confounding that can be attributed to shared environmental factors (i.e., household economic disadvantage, maternal education, etc.). Of course, this removes the need to specify traits/environments that are shared by twins within a twin pair and include them in the model (e.g., age, sex, race, household

102 income, maternal education, female-headed household, etc.). Prior research linking nutritional factors to behavioral problems has primarily used nonexperimental research designs, which are often plagued by residual confounding due to the inability to properly control for a host of familial factors that are shared among siblings in the same household, especially those pertaining to maternal traits. Second, DF analysis is capable of modeling the proportion of the variance in the outcome of interest that can be attributed to genetic influences, which strengthens causal inferences about the influence of the nonshared environmental factors that are examined simultaneously (e.g., sibling differences in breastfeeding duration, dietary habits). DF analysis provides a more rigorous test of environmental influences than other observational research because of its ability to a) distinguish between shared and nonshared environmental influences and b) test whether specific nonshared environmental effects are robust to the effects of genetic factors.

Although the current study is not experimental, it answers the call of Shelton and colleagues (2011) and other researchers (see D’Onofrio et al., 2013), through the use of the

Defries-Fulker technique, to examine whether siblings who are discordant in their nutritional behaviors during infancy and early childhood are also discordant in their degree of externalizing behavioral problems during their kindergarten school year, net of genetic and shared environmental influences. Genetically informed tests that employ sibling differences as a way to tap the influence of nonshared nutritional factors on subsequent behavior during early childhood are virtually nonexistent (except see a unique prenatal cross-fostering research design by Shelton et al., 2011). This oversight in the literature is one of the impetuses for the current study.

103 The DF equation has been revised since it was originally postulated by DeFries and

Fulker (1985, 1988) in order to be fit for use among samples drawn from the general population

(Rodgers, Rowe, & Li, 1994). The revised equation is depicted as follows:

Equation 1: K1 = b0 + b1K2 + b2R + b3(R * K2) + e,

The exact interpretation of the above equation will vary, depending on the outcome variable being examined. In the current study, externalizing behavior is most frequently examined as the outcome variable, although, in some analyses, other variables are employed as outcome variables

(e.g., low attachment security, low diet quality). For the sake of simplicity, the above example of a DF equation, and all of its subsequent forms, will be interpreted using externalizing behavior as the outcome variable. With that in mind, K1 in the above equation represents the externalizing behavior score (i.e., the outcome variable) for one of the twins being analyzed, K2 represents their cotwin’s externalizing behavior score, R is an indicator of the genetic similarity between the kinship pair (1 for MZ twin pairs and .5 for DZ twin pairs), and R * K2 is an interaction term that multiplies the cotwin’s externalizing behavior score by their degree of genetic similarity with their twin. Moreover, b0 represents the constant, b1 represents the proportion of the variance in externalizing behavior that is explained by shared environmental influences, b2 is not interpreted in the DF model, and b3 is the proportion of the variance in externalizing behavior that is explained by genetic influences. The error term (e) encompasses the effects of the nonshared environment on externalizing behavior and error.8

Recently, Rodgers and Kohler (2005) proposed an improvement to equation 1 that only slightly alters its form. The new equation is depicted as follows:

Equation 2: K1 = b0 + b1(K2 – Km) + b2[R * (K2 - Km)] + e,

8 In some follow-up analyses, K1 and K2 represent the twins’ scores on other variables, including low attachment security and low diet quality.

104 In this equation, K1, K2, R and e have the same significance as they do in equation 1. However, this equation includes the term Km, which represents the mean value of K2 (or, in this study, the mean externalizing behavior score of the cotwins). Therefore, the parenthetical statement K2 –

Km signifies that K2 is mean-centered in this equation. Just as was the case in equation 1, b0 represents the constant and b1 represents the proportion of the variance in externalizing behavior that is explained by shared environmental influences. However, in this updated equation, b2

(instead of b3) is interpreted as the proportion of the variance in externalizing behavior that is explained by genetic influences.

The coefficients in the above equation do not reveal the effect of any particular gene or shared environment on externalizing behavior precisely because the coefficients signify latent factors. Nevertheless, equation 2 can be altered slightly to allow for the inclusion of specific nonshared environments of interest. In the current study, I make use of the following equation in order to examine a number of nonshared environments related to breastfeeding duration and early childhood nutritional behaviors. Doing so permits me to determine whether these nonshared environments have a significant influence on externalizing behavior, net of genetic and shared environmental factors. The DF equation that allows researchers to include specific nonshared sources of variance is depicted as follows:

Equation 3: K1 = b0 + b1(K2 – Km) + b2[R * (K2 - Km)] + b3ENVDIF + e,

Equation 3 is almost an exact replication of equation 2. The only difference is the term

ENVDIF. ENVDIF represents the difference score that is created when one twin’s score on a variable is subtracted from their cotwin’s score on the same variable. In the current study, difference scores are calculated for each of the nutritional variables, as well as low attachment security, in order to determine if sibling differences in these variables predict differences in

105 9 externalizing behavior, net of genetic and shared environmental influences. Importantly, b3 in equation 3 does not represent a latent factor, but instead represents a regression coefficient, and needs to be interpreted as such (e.g., using critical t-values, p-values, etc.).

A series of DF models was estimated in the present study. The first model employs the baseline DF equation (Equation 2) in order to ascertain the proportion of the variance in externalizing behavior that is due to genetic, shared environmental, and nonshared environmental factors. Subsequent models employ the formula displayed in Equation 3, which allows for the introduction of the breastfeeding and nutritional variables as nonshared sources of variance by including them as difference scores (ENVDIF) in the equation. The goal of these analyses is to determine whether differences between the twins in breastfeeding duration and early childhood dietary patterns significantly contribute to differences in their externalizing behavior scores during kindergarten, independent of genetic and shared environmental influences.10

In order to maximize the information available on twin pairs in the ECLS-B, and in line with prior research (Beaver et al., 2009; Haynie & McHugh, 2003; Kohler & Rodgers, 2001;

Rodgers, Buster, & Rowe, 2001), twins were double-entered. Double entering is the most frequent choice among researchers when using the augmented DF equations. Rodgers and colleagues (2001) argue that double entering is the correct approach when the specification of which siblings represent K1 and which siblings represent K2 is arbitrary, which is the case in the current study.11 Double entering allows for each twin to be both the independent and dependent variables in the DF analysis. Despite this advantage, double entering violates the assumption of the independence of observations (since the same observations are repeated twice). Violation of

9 Again, in some follow-up analyses, K1 and K2 represent the twins’ scores on other variables, including low attachment security and low diet quality. 10 Again, in some follow-up analyses, K1 and K2 represent the twins’ scores on other variables, including low attachment security and low diet quality. 11 Additionally, results showed no substantive variation using the alternative approach of single entering the data.

106 this assumption results in deflated standard errors, which biases tests of statistical significance.

In line with prior research (Beaver et al., 2009; Haynie & McHugh, 2003), I corrected for this bias by employing Huber-White standard errors, which takes account of the clustering of observations when estimating the statistical significance of the results. In this study, DF analysis was used to explore the link between nutritional factors, low attachment security, and externalizing behavior (i.e., research questions 2, 3, 6, 7, and 10).

5.4.3 Cotwin Interactive Analysis

In addition to DF analysis, another genetically informative technique was used to test for the presence of various interactions between the breastfeeding/nutritional measures and a) genetic risk, b) low attachment security and, in the breastfeeding section, c) low birth weight, while controlling for level of genetic risk. Upon establishing the relevance of genetic influences in the emergence of externalizing behavior using DF analysis, as well as the relevance of breastfeeding and dietary patterns as seemingly significant nonshared environmental influences,

I consider the possibility that the impact of a short duration of breastfeeding and poor childhood nutrition on externalizing behaviors might be heightened in the presence of genetic, perinatal, and/or socioemotional risk factors (e.g., low attachment security). To test these hypotheses, I created multiplicative interaction terms between the breastfeeding measures and genetic risk, low attachment security, and low birth weight. Subsequently, I created multiplicative interaction terms between the early childhood nutritional measures and genetic risk and low attachment security. Interaction terms using the breastfeeding measures and the early childhood diet measures were also created in order to examine whether the effects of nutritional factors on externalizing behavior at one life stage were conditioned by nutritional factors at the other life

107 stage. As was the case in the DF analysis, Huber-White standard errors were used to adjust for the bias in standard errors due to the use of the same-sex twin sample. Furthermore, the components of each interaction term were standardized prior to multiplication (Jaccard, Wan, &

Turrisi, 1990). In the current study, cotwin interactive analysis was employed to test an array of biosocial interactions between nutrition, perinatal, genetic, and socioemotional processes

(research questions 4, 8, and 11).

108 CHAPTER 6

RESULTS

The results of the current study are presented in three sections labeled 6.1, 6.2, and 6.3.

In the first section, the results of the analyses pertaining to nutrition during infancy (i.e., breastfeeding) will be presented (i.e., research questions 1-4). In the second section, the results of the analyses pertaining to nutrition during early childhood (i.e., low diet quality) will be presented (i.e., research questions 5-8). In the third and final section, the results of the analyses pertaining to the interplay of nutritional factors across infancy and early childhood will be presented (i.e., research questions 9-11). Within each of these three sections, results will be arranged by research question. For additional details regarding the methods used, please see

Chapter 5 (METHODS).

6.1 Results: Breastfeeding Analyses

The breastfeeding analyses employ each of the analytical techniques described in chapter

5 (METHODS) in an effort to answer the following four research questions:

1. In what ways do mothers and children who breastfeed for shorter durations differ from

mothers and children who breastfeed for longer durations?

2. Does a short duration of breastfeeding significantly increase the risk of low attachment

security during toddlerhood and/or externalizing behavior problems during kindergarten,

independent of familial and genetic influences?

3. Is the relationship between short duration of breastfeeding and externalizing behavior

problems during kindergarten explained by low attachment security during toddlerhood

and/or low birth weight?

109 4. Is the influence of breastfeeding duration on kindergarten externalizing behavior

moderated by genetic risk, low attachment security, and/or low birth weight?

6.1.1 Research Question 1

The first research question asks: In what ways do mothers and children who breastfeed for shorter durations differ from mothers and children who breastfeed for longer durations?

As noted previously, table 6.1 displays the descriptive statistics of all variables included in the breastfeeding analyses for the same-sex twin sample. The univariate results reveal that the average duration of breastfeeding in the sample is approximately 3 months, with a standard deviation of 4.59 months. The sample is exactly 50% male, 42% nonwhite, and with an average age of 69.46 months (or approximately 5 years and 9 months) at waves 4/5. Mothers were, on average, about 30 years old at wave 1 and roughly 18% of them were raising their children without a partner/husband in the home. The average score on the low household income variable at wave 1 was approximately 5 (or $35,000 - $40,000 per year) and the average score on low maternal education was also approximately 5 (which equates to some college).

Table 6.2 displays the bivariate correlations between all variables used in the breastfeeding analyses. Notably, the breastfeeding variables (e.g., short duration of breastfeeding, breastfed less than 6 months) were positively and significantly correlated with externalizing behavior during kindergarten, as well as a number of covariates, including low maternal education and female-headed household. Measures of breastfeeding that do not take exclusivity into account appear to be the most highly and consistently correlated with the covariates. For example, a shorter duration of breastfeeding, regardless of exclusivity, is significantly and positively correlated with all covariates except for three: the sex of the child, the age of the child, and the age of the mother. Younger mothers, however, tended to breastfeed

110 for shorter durations, yielding a significant, negative correlation. The table also indicates that nonwhite children, as well as those who reside in a low income household, tend to breastfeed for shorter durations. Low birth weight children, children with low attachment security, and children with high level of genetic risk for externalizing behavior, moreover, are also more likely to breastfeed for shorter durations. In regards to mothers, those who live without a spouse/partner, have lower levels of education, experience significant postpartum depression, and are less involved with their infant are also more likely to breastfeed for shorter durations.

Despite the overall pattern of significant correlations between the covariates and the breastfeeding variables, correlations were noticeably weaker (and sometimes non-significant) between many of the covariates and the Not Exclusively Breastfed measure. Thus, it appears to be more challenging to predict the mother-infant pairs who are most likely to fail to breastfeed exclusively for 6 months, since race, low household income, postpartum depression, low parental involvement, and low attachment security are not significantly correlated with the exclusivity of breastfeeding.

Table 6.3 displays the results of several difference of means t tests which compare child, maternal, and household covariates across the 6-month breastfeeding threshold. Specifically, the table illustrates 1) the average score on each covariate for each of the two groups distinguished by the breastfeeding threshold of 6 months, 2) the difference between the two means on each covariate, 3) the t-value that corresponds to the difference of means and 4) the statistical significance of the difference of means as denoted by the obtained p-value.12 The difference of

12 In the case of maternal/household covariates, the sample used to determine significance was the sample of households, not individual children, as there were two children per household in the same-sex twin sample. To determine group classification based on breastfeeding duration, one child was randomly chosen from each household to create the household sample upon which these t tests are based. However, in the case of child covariates, which were measured separately for each child within each household, the sample of individual children (not households) was used to determine significance.

111 means t tests yielded several interesting findings regarding child, maternal, and household characteristics of mother-infant pairs who breastfed for less than 6 months, relative to those who breastfed for at least 6 months.

First, Table 6.3 reveals that children with certain traits were less likely to reach the 6- month threshold. For example, 52% of the children who breastfed for less than 6 months were male, whereas only 42% of the children who breastfed for 6 months or longer were male (t =

2.48; p = .01), a difference of 10 percentage points. Additionally, 44% of the children who breastfed for less than 6 months were nonwhite, whereas only 33% of the children who breastfed for 6 months or longer were nonwhite (t = 2.91; p = .00), a difference of 11 percentage points.

Thus, on average, male and nonwhite children were significantly more likely to breastfeed for less than 6 months, and were underrepresented among children who breastfed for at least 6 months. Results also reveal that, relative to children who breastfed for 6 months or longer, children who breastfed for less than 6 months tended to 1) be characterized by lower attachment security (t = 4.12; p = .00), 2) be born low birth weight (t = 3.36; p = .00), and 3) possess a greater degree of genetic risk for externalizing behavior (t = 3.61; p = .00). To illustrate, 63% of children who breastfed for less than 6 months were born low birth weight, whereas only 50% of children who breastfed for 6 months or more were born low birth weight.

Maternal and household characteristics were also found to vary significantly across the 6 month breastfeeding threshold. For instance, the < 6 month group scored, on average, a full unit higher on low household income and nearly a full unit higher (.82) on low maternal education.

These differences imply that children who belong to the < 6 month group will, on average, reside in households making anywhere between $5,000 and $10,000 less in annual income relative to the >= 6 month group. Furthermore, these findings suggest that, while the average level of

112 education of mothers who breastfed for 6 months or more is a bachelor’s degree, mothers who do not reach the 6-month mark, on average, attend but do not graduate college. As detected in the bivariate correlation analysis, mothers who do not reach the 6-month breastfeeding threshold tend to be significantly younger than those that do. To be precise, mothers who breastfeed for less than 6 months are approximately 3 years younger, on average, than those that breastfeed long term (29.38 v. 32.14). Perhaps most notably, hardly any mothers who breastfed long term

(i.e., 6 months or longer) were running a household on their own (.05, or 5%). Conversely, more than a fifth of mothers who either never initiated breastfeeding or breastfed for short durations were running a household on their own (.21, or 21%). Finally, mothers who breastfed for less than 6 months, on average, were significantly less involved with their infant (t = 3.54; p = .00) and more depressed (t = 2.32; p=.02) at wave 1 than mothers who breastfed long term.

Ultimately, the results displayed in table 6.3 indicate that a short duration of breastfeeding does not occur in a vacuum, but co-occurs with various contextual risk factors pertaining to the child, the mother, and the household.

Table 6.4 displays the results of additional difference of means t tests. However, instead of illustrating child, maternal, and household covariates across the 6-month breastfeeding threshold, table 6.4 displays the results of t tests that compare key covariates across mother- infant pairs who breastfed exclusively for long durations (>= 6 months) and mother-infant pairs who did not breastfeed exclusively for long durations. Interestingly, the pattern of group differences across the covariates is not as striking as in table 6.3. Of the 12 measured covariates, only three are statistically significant at the .05 level: sex of the child, low maternal education, and female-headed household (although two additional covariates are nearly significant). To illustrate, while over half of the children who were not exclusively breastfed for at least 6 months

113 were male (51%), only 33% of children who were exclusively breastfed for 6 months or more were male (a difference of 18 percentage points). Similarly, mothers who did not exclusively breastfeed for long durations scored 1.33 units higher on the low maternal education variable than mothers who did breastfeed exclusively for long durations. One of the most striking differences between groups pertains to the rate of single motherhood within each group. While

18% of mothers who failed to exclusively breastfeed long term were single mothers, none of the mothers who engaged in long-term, exclusive breastfeeding were single mothers. Thus, female- headed households were not represented among the small subset of the sample who managed to engage in exclusive breastfeeding for 6 months.

Although there was an overall tendency for mothers who did not participate in long-term, exclusive breastfeeding to be younger than those that did (difference = -2.83), this difference was only of borderline significance (p = .06). Similarly, children who were not exclusively breastfed for at least 6 months tended to possess a greater degree of genetic risk as well as display a lower security of attachment when compared to exclusively breastfed children. Again, however, these differences did not quite reach statistical significance (p = .06 and p = .08). By way of comparison, there was little evidence to suggest that nonwhite children and those born low birth weight were differentially represented between the groups distinguished by exclusivity (see table

6.4), even though such children were overrepresented in the short-term breastfeeding group that did not account for exclusivity (see table 6.3). Also, tables 6.4 reveals that mothers who engage in long-term, exclusive breastfeeding are not significantly different from mothers who fail to do so in regards to low parental involvement, postpartum depression, and low household income, all of which predicted whether mother-child pairs reached the 6-month threshold irrespective of exclusivity (see table 6.3). As intimated in the bivariate correlations, the t tests reveal an

114 interesting paradox: while the 6-month threshold is a robust indicator of multiple dimensions of maternal, child, and household risk, it is nonetheless considerably more difficult to pinpoint a cluster of traits that uniquely identify mother-infant dyads who participate in long-term, exclusive breastfeeding.

6.1.2 Research Question 2

The second research question asks: Does a short duration of breastfeeding significantly increase the risk of low attachment security during toddlerhood and/or externalizing behavior problems during kindergarten, independent of familial and genetic influences?

As illustrated in the bivariate analyses of the same-sex twin sample, the duration of breastfeeding is closely related to a number of important environments that same-sex twins within the same household would be equally exposed to, namely low household income, maternal education, female-headed household, age of mother, and postpartum depression of the mother. In a similar manner, same-sex twins share traits that are also correlated with breastfeeding duration, namely sex and race. Thus, in an effort to rule out residual environmental confounding (due to poor measurement and/or omitted variable bias) as well as potential genetic confounding, I proceed by addressing research question 2: “Does a short duration of breastfeeding significantly increase the risk of low attachment security during toddlerhood and/or externalizing behavior problems during kindergarten, independent of familial and genetic influences?”. This research question will be addressed using DF analysis (see

Chapter 5 for more details).

The use of the same-sex twin sample facilitates a genetically informative analysis for two reasons. First, the sibling sample allows for comparisons of children within the same family, which is essential in order to rule out key confounders that constitute shared environments

115 between siblings (e.g., postpartum depression of the mother, low maternal education, low household income). On a related note, comparing same-sex twins eliminates age, sex and race as explanations of sibling differences in the outcomes of interest. Second, using same-sex twins with differing degrees of genetic similarity (i.e., MZ, DZ) allows for a modeling strategy that takes genetic factors into account when estimating the effects of environmental factors.

Essentially, the DF method allows for the specification of nonshared environmental influences

(e.g., sibling difference scores on nutritional factors) on an outcome of interest (e.g., externalizing behavior, low attachment security), while effectively controlling for the effects of genetic and shared environmental factors. This is done by estimating the proportion of the variance in the outcome that is due to shared environmental effects and genetic effects. To the extent that a portion of the variance in the outcome remains unexplained by either shared environmental or genetic factors, specific nonshared environmental factors can be modeled as potentially important predictors of the outcome. In the current study, child nutritional factors

(and, in some cases, low attachment security) are modeled as nonshared environments (i.e., difference scores between twins within a twin pair). Specific shared environments (e.g., low maternal education, female-headed household) are incapable of confounding the results of these models, and thus need not be explicitly modeled.

Table 6.5 displays the descriptive statistics of the variables used in the DF analyses pertaining to breastfeeding practices. Although there was a great deal of concordance in breastfeeding variables among twins, there was still a notable degree of discordance (roughly

12% of twins were discordant). As illustrated in the table, some sibling differences in breastfeeding duration were quite variable, with twins within a twin pair differing in breastfeeding duration by as much as 20 months. Difference scores on low attachment security

116 also displayed a wide range, with the majority of twins within each pair displaying some difference from their cotwin in their degree of attachment security. Specifically, while the observed, untransformed scores for the twins ranged from -.98 to .86 (an absolute range of 1.84), the largest twin difference in low attachment security was 1.52 units, which constitutes approximately 83% of the possible range of untransformed scores. The implication of this finding is that the current sample includes a number of securely attached twins who have relatively insecurely attached cotwins.

I now turn to the DF portion of the breastfeeding analyses. The results of the first DF model are displayed in table 6.6. Model 1 of table 6.6 contains the baseline model, which does not estimate the specific environmental effects of interest, but only estimates the latent factors

(i.e., shared environment and heritability). This model indicates that approximately 83% of the variance in externalizing behavior problems during kindergarten can be attributed to genetic factors (p < .01). Because the estimated shared environmental effect was .00 (and nonsignificant), the remaining portion of the variance (17%) can be attributed to non-shared environmental factors and error. Models 2 through 4 of table 6.6 extend model 1 by including specific measures of breastfeeding as nonshared environmental influences. The coefficients for the nonshared sources of variance can be interpreted just as OLS regression coefficients are interpreted, realizing, however, that the unit of analysis is twin pairs (not individuals). To be precise, in the current analysis, these coefficients represent the average increase in externalizing behavior for every one-unit increase in the nonshared source of variance (e.g., short duration of breastfeeding), relative to one’s cotwin.

Model 2 of table 6.6 specifically examines whether differences between cotwins in their duration of breastfeeding predict their relative likelihood of externalizing behavioral problems

117 during kindergarten. In this model, breastfeeding is operationalized as a continuous construct that is measured in months, with higher scores indicating fewer months of breastfeeding (relative to one’s cotwin). The results of model 2 indicate that breastfeeding does not have a significant dose-response effect on externalizing behavioral problems when genetic and shared environmental effects are taken into account (i.e., additional months of breastfeeding do not correspond to lower externalizing behavior scores). Put differently, twins who terminate breastfeeding earlier than their cotwins do not appear to be at greater risk of developing externalizing behavioral problems by kindergarten.

Nevertheless, the results of model 2 must be qualified in light of the results depicted in model 3. In model 3, breastfeeding duration is modeled as a dichotomous variable to test for the possibility of threshold effects at the 6-month mark. That is, although small differences in the extent of breastfeeding exposure between cotwins do not appear to influence their relative likelihood of externalizing behavior problems during kindergarten, it is important to consider the possibility that significant effects may emerge if a twin attains or surpasses an important breastfeeding threshold (e.g., 6 months), but their cotwin fails to do so. The results of model 3 suggest that, independent of genetic and shared environmental influences, a significant threshold effect of breastfeeding on externalizing behavior exists at the 6-month mark, despite no evidence of a dose-response effect in model 2 (p < .01). Put differently, among twins who are discordant in their length of breastfeeding, twins who breastfed for less than 6 months exhibit significantly higher levels of externalizing behavior than cotwins who breastfed for 6 months or more. Thus, when discordance in the length of breastfeeding is modeled by designating a 6-month threshold, significant effects emerge; however, when differences are modeled irrespective of the threshold,

118 breastfeeding appears to be irrelevant to the development of externalizing behavior (see figure

6.1 for an illustration of this finding).

Model 4 of table 6.6 takes the analysis one step further and explores whether failing to exclusively breastfeed for 6 months also places offspring at risk for externalizing behavior. The results of model 4 indicate that children who are not exclusively breastfed for at least 6 months are not at an increased of externalizing behavior problems during kindergarten. More specifically, twins who failed to breastfeed exclusively for at least 6 months exhibited similar levels of externalizing behavior during kindergarten as twins who breastfed exclusively for 6 months or more. Thus, while model 3 detected a significant threshold effect of breastfeeding on externalizing behavior, this effect seems to disappear upon accounting for the exclusivity of breastfeeding. That is, once genetic and shared environmental factors are controlled, it appears that the externalizing behavior of children who participate in long-term, exclusive breastfeeding is not significantly higher/worse than the externalizing behavior of children who fail to engage in long-term, exclusive breastfeeding. These null results must be qualified, however, in light of the emergence of a significant threshold effect at 6 months on externalizing behavior, regardless of exclusivity. It appears, then, that it is the failure to reach a particular duration of breastfeeding (6 months), and not whether such breastfeeding is exclusive, that has significant implications for the externalizing behavior of offspring during kindergarten.

Table 6.7 employs the same DF technique that was employed in the previous analyses

(the results of which were displayed in table 6.6). However, in this case, low attachment security during toddlerhood, which was measured separately for each twin, is modeled as the outcome variable of interest. As discussed previously (see Chapter 4), a number of scholars have argued that breastfeeding may be acting as a proxy for enhanced infant-mother bonding and associated

119 levels of attachment security (Britton et al., 2006; Gribble, 2006; Tharner et al., 2012). Because the process of breastfeeding and mother-infant attachment may be closely intertwined, it is worthwhile to appropriately tease out whether breastfeeding, attachment security, or both are ultimately influencing the degree of externalizing behavior exhibited by offspring during kindergarten. Importantly, however, the notion that breastfeeding and attachment security are so closely intertwined is, in many ways, a taken-for-granted assumption that may not be empirically sound (see Else-Quest et al., 2003; Jansen et al., 2008; Papp, 2013; Schulze & Carlisle, 2010). In light of the current body of contradictory results, the present study seeks to determine the validity of the frequently held assumption that a short duration of breastfeeding will interfere with, and weaken, the security of mother-child attachment during toddlerhood. Doing so will ultimately allow for a more rigorous and methodologically sound test of the relationship between breastfeeding practices and externalizing behavior in offspring.

The results of the DF model pertaining to the relationship between breastfeeding practices and low attachment security during toddlerhood are displayed in table 6.7. As was the case in table 6.6, model 1 of table 6.7 contains the baseline model, which only estimates the latent factors (i.e., shared environment and heritability). This model indicates that approximately

52% of the variance in low attachment security during toddlerhood can be attributed to shared environmental factors (p < .01). This suggests that just over half of the variance in low attachment security among the twins in the sample can be explained by environments that twins within a twin pair share. Conversely, genetic factors did not explain a significant portion of the variance in low attachment security during toddlerhood. To be precise, although the estimated slope for heritability was .19 (b = .19, p > .05), it should be interpreted as b = 0, since the 95% confidence interval surrounding the estimate includes the value of 0. As a result, 0% of the

120 variance in low attachment security can be attributed to genetic factors. This implies that the remaining portion of the variance (.48, or 48%) can be attributed to the nonshared environment and error.

Models 2 through 4 of table 6.7 build upon the first model by including specific measures of breastfeeding as nonshared environmental influences. The coefficients corresponding to each breastfeeding measure represent the average increase in low attachment security for every one- unit increase in the nonshared source of variance (e.g., short duration of breastfeeding), relative to one’s cotwin. Model 2 of table 6.7 explores whether differences between twins in their duration of breastfeeding predict their relative likelihood of low attachment security during toddlerhood. As was the case in model 2 of table 6.6, breastfeeding is operationalized as a continuous construct that is measured in months, with higher scores indicating fewer months of breastfeeding (relative to one’s cotwin). The results of model 2 of table 6.7 suggest that twins who terminate breastfeeding earlier than their cotwins develop significantly lower levels of attachment security than their cotwins. In other words, there is evidence of a dose-response effect of breastfeeding on low attachment security, with a lower dose of breastfeeding (i.e., in units of months) corresponding to a lower level of attachment security.13 These results appear to support the arguments of scholars like Gribble (2006), who suggested that there are certain non- nutritional benefits of breastfeeding (e.g., greater responsiveness and proximity of the mother to her baby) that likely facilitate greater attachment security. Importantly, these results pertain to within family variation and are genetically informative, and thus cannot be confounded due to genetic risk and/or due to risk factors shared by twins living in the same household. The finding

13 Additional DF analyses were performed to examine the relationship between breastfeeding and each of the nine hotspots that inform the low attachment security measure. The results revealed that a shorter duration of breastfeeding impacted seven of the hotspots, including is upset by separation (b = .02; p = .03), enjoys company (b = -.02; p = .00), seeks attention (b = .02; p = .00), is independent (b = -.02; p = .01), is demanding/angry (b = .03; p = .00), is cooperative (b = -.03; p = .00), and is comfortably cuddly (b = -.02; p = .03).

121 detected in model 2 provides some initial evidence that breastfeeding is indeed a dyadic process, and that even the mother of a pair of same-sex twins can interact quite differently with each of her twins by toddlerhood, and that this process is predicted by differences in breastfeeding duration.

The results of model 2, however, must be tempered in light of the results displayed in model 3. To be precise, when breastfeeding is modeled as a threshold effect at the 6-month mark, twins who breastfed for less than 6 months did not exhibit a significantly lower level of attachment security during toddlerhood than twins who breastfed for 6 months or more. In other words, the results of model 3 reveal no evidence of a 6-month threshold effect of breastfeeding on low attachment security.14 The results of model 4 are consistent with those of model 3: twins who did not engage in long-term, exclusive breastfeeding were not at significantly greater risk of developing a low level of attachment security than twins who engaged in long-term, exclusive breastfeeding. Thus, whether exclusivity is considered or not, this study finds no evidence that reaching or surpassing the 6-month threshold of breastfeeding provides any additional benefits in terms of overall toddler attachment security beyond the month-to-month benefits detected in model 2.15

In sum, certain breastfeeding practices appear to influence both externalizing behavior during kindergarten as well as low attachment security during toddlerhood. However, the precise impact of breastfeeding is distinct in each case. In the case of externalizing behavior, there is only evidence of a threshold effect at the 6-month mark, yet no evidence of a dose- response, or month-to-month, effect. In the case of low attachment security, there is evidence of

14 The findings were similar when individual hotspots were examined as outcomes. However, breastfeeding less than six months did significantly predict scores on is demanding/angry (b = .16; p = .00), seeks attention (b = .14; p = .00), and is upset by separation (b = .11; p = .02). 15 The findings were similarly null when individual hotspots were examined as outcomes.

122 a dose-response effect of breastfeeding, yet no evidence of a threshold effect at the 6-month mark. Interestingly, neither the degree of externalizing behavior nor the level of attachment security were significantly influenced by the failure to exclusively breastfeed for at least 6 months. Consequently, while exclusivity has little bearing on these outcomes, the exact duration of breastfeeding does. Monthly decreases in the duration of breastfeeding appear to interfere with attachment security. However, reductions in the duration of breastfeeding would not be expected to have an impact on externalizing behavior during kindergarten, unless such reductions dropped below the 6-month threshold (see figure 6.1 for an illustration).

6.1.3 Research Question 3

The third research question asks: Is the relationship between short duration of breastfeeding and externalizing behavior problems during kindergarten explained by low attachment security during toddlerhood and/or low birth weight?

In light of the findings linking different measures of breastfeeding practices to both the externalizing behavior and the degree of attachment security of offspring, the next logical question is whether (and to what extent) low attachment security explains the link between breastfeeding less than 6 months and externalizing behavior during kindergarten. Table 6.8 of displays the results of a DF analysis of the shared environment, heritability, and low attachment security as predictors of externalizing behavior during kindergarten. In this model, low attachment security (relative to one’s cotwin) is posited as a key nonshared source of variance that might explain variation in externalizing behavior between twins within a twin pair, net of shared environmental and genetic influences.16 As detected previously, model 1 of table 6.8

16 It might be argued that behavioral indicators of attachment security and the items that comprise the externalizing behavior measure are similar in nature. Prior research on attachment security and its measurement has noted that it is not merely the presence of certain behaviors that determine the level of attachment security, but rather it is the organization of such behaviors and the specific context in which these behaviors occur (Walter & Deane, 1985).

123 reveals that 83% of the variance in externalizing behavior during kindergarten is explained by genetic factors, with 0% of the variance being explained by the shared environment. Model 2 displays the results when the baseline model is expanded to include twin differences scores in low attachment security. The results reveal that twins who exhibit lower levels of attachment security during toddlerhood (relative to their cotwins) engage in a significantly greater number of externalizing behaviors during kindergarten (relative to their cotwin).17

Thus, in conjunction with the results of prior models, these results suggest that attachment security, as well as breastfeeding practices, influence the degree of externalizing behavior during kindergarten. Furthermore, month-to-month differences in breastfeeding between twins within a twin pair are significantly related to their level of attachment security during toddlerhood. In light of these findings, it is worthwhile to explore whether the effects of breastfeeding practices on externalizing behavior persist once the level of attachment security during toddlerhood is accounted for in the model. Table 6.9 reveals the results of these analyses.

Model 1 of table 6.9 reveals the baseline model, which exactly mirrors the results depicted in model 1 of table 6.8. The more pertinent results concerning this research question are found in models 2 through 4 of table 6.9. Model 2 reveals that, while low attachment security predicts externalizing behavior during kindergarten, month-to-month changes in

Most importantly, attachment security is rooted in the mother-child relationship and the way that children respond to the presence or absence of various environmental stimuli associated with the mother. Although many behavioral indicators are used in the TAS45, the tool is nonetheless employed as a means of determining the degree of socioemotional connection between the mother and the child. Additionally, factor analyses revealed that the items most characteristic of low attachment security [e.g., comfortably cuddly (reverse-coded), cooperative (reverse- coded), demanding/angry] and the measure of externalizing behavior load onto distinct factors. 17 Additional DF analyses were performed to examine the relationship between each of the nine hotspots that comprise the attachment security measure and externalizing behavior. The results revealed that several hotspots impacted externalizing behavior, including is demanding/angry (b = .18; p = .00), is cooperative (b = -.23; p = .00), and is comfortably cuddly (b = -.23; p = .00). Furthermore, is moody/unusual/unsure was marginally significant (b = .08; p = .10). Importantly, the hotspots that emerged as significant predictors of externalizing behavior are integral to the attachment measure (see Figures B.1 and B.2). Ultimately, insecurely attached children fail to regularly engage with their parent(s) in a warm, affectionate, and cooperative manner, but instead frequently engage with their parent(s) in an angry, moody, and demanding way. For more information, see Appendix B.

124 breastfeeding duration do not predict externalizing behavior independent of low attachment security during toddlerhood. These results, however, are not uniform across the different measures of breastfeeding practices, as illustrated in model 3. Specifically, the results of model

3 indicate that both breastfeeding less than 6 months and low attachment security have significant, direct effects on externalizing behavior during kindergarten. In short, the results suggest that, while breastfeeding duration and attachment security may be related, threshold effects of breastfeeding at the 6-month mark on externalizing behavior are not explained by low attachment security. Finally, model 4 of table 6.9 reveals that failing to exclusively breastfeed for 6 months does not significantly impact externalizing behavior during kindergarten, independent of low attachment security.

The results of the models as a whole indicate that, of the breastfeeding measures, only the breastfeeding threshold measure has a significant, direct effect on externalizing behavior, which is independent of the significant influence of low attachment security. Nonetheless, as revealed in previous models, month-to-month differences in breastfeeding duration significantly impact the degree of attachment security (see table 6.7, model 2), which in turn influences externalizing behavior scores during kindergarten (see table 6.8, model 2). Consequently, while threshold effects of breastfeeding exist independent of attachment security, there is also evidence of a completely indirect dose-response effect of breastfeeding on externalizing behavior, since differences in months of breastfeeding predict differences in the level of attachment security, and the relative level of attachment security predicts the relative degree of externalizing behavior.

Finally, it could be argued that twin differences in breastfeeding duration are merely acting as proxies for their degree of perinatal risk. For example, research has indicated that low birth weight infants are less likely to initiate breastfeeding and/or persist for long durations (see

125 Smith et al., 2003). Perhaps most notable is the significantly decreased rate of breastfeeding for

6 months or more among low birth weight infants. For instance, Flacking, Nyqvist, Ewald, &

Wallin (2003) found that the rate of breastfeeding for 6 months or more among their Swedish sample of low birth weight infants was less than half the rate of the county population. It is reasonable, therefore, to suggest that differences in birth weight status between twins may confound the breastfeeding-externalizing link, particularly since low birth weight has been shown to increase the risk of developing externalizing behavior problems at subsequent life stages (Breslau & Chilcoat, 2000; Wadsby et al., 2014). Table 6.10 displays the results of several models which test whether any effect of breastfeeding practices on externalizing behavior is explained by low birth weight status. Importantly, the results reveal that low birth weight did not significantly impact externalizing behavior during kindergarten, independent of shared environmental and genetic influences. Twin differences in low birth weight status, therefore, did not significantly predict their relative likelihood of exhibiting externalizing behavior problems.

As a result, low birth weight was incapable of confounding the direct relationship between breastfeeding for less than 6 months and externalizing behavior detected in model 3.18

In sum, the link between breastfeeding less than 6 months and externalizing behavior during kindergarten appears to be robust to both low attachment security and low birth weight.

Although low birth weight was not significantly related to externalizing behavior net of genetic and shared environmental factors, low attachment security remains an important, independent predictor of externalizing behavior. The results of several DF models in conjunction also suggest that, apart from the direct, 6-month threshold effect of breastfeeding on externalizing

18 A sensitivity analyses was also conducting using a continuous measure of birth weight. The results of this analysis yielded substantively identical results.

126 behavior, month-to-month reductions in breastfeeding duration may have less obvious, entirely indirect effects on externalizing behavior by decreasing the level of attachment security.

6.1.4 Research Question 4

The fourth research question asks: Is the influence of breastfeeding duration on kindergarten externalizing behavior moderated by genetic risk, low attachment security, and/or low birth weight?

In addition to exploring whether breastfeeding practices significantly influenced kindergarten externalizing behaviors, I also tested whether the influence of a short duration of breastfeeding on externalizing behavior might be conditioned by genetic risk, low attachment security, and/or low birth weight. Although these genetic, perinatal, and socioemotional factors did not explain or confound the relationship between breastfeeding for less than 6 months and externalizing behavior during kindergarten, they may exert important interactive effects on externalizing behavior in conjunction with breastfeeding practices. Put differently, a short duration of breastfeeding might be more or less relevant to offspring externalizing behavior, depending on their level of risk across genetic, perinatal, and socioemotional domains. Table

6.11 specifically displays the models that include interactions between the breastfeeding measures and genetic risk predicting externalizing behavior during kindergarten. In particular, model 1 interacts the continuous measure of a short duration of breastfeeding with genetic risk, model 2 interacts the dichotomous, 6-month threshold measure with genetic risk, and model 3 interacts the exclusivity measure (i.e., Not Exclusively Breastfed) with genetic risk.19

The results are consistent across each of the three models displayed in table 6.11. To be precise, the results indicate that, whether breastfeeding is measured continuously or

19 Multicollinearity analyses revealed acceptable levels of collinearity for all regression models (the variance inflation factors range from 1 to 2, with tolerance statistics ranging from .50 to .96).

127 dichotomously, exclusively or nonexclusively, genetic risk appears to condition the effect of a short duration of breastfeeding on the externalizing behavior of offspring during kindergarten. A shorter duration of breastfeeding, therefore, is particularly likely to amplify childhood externalizing behavior for those children who a have higher level of genetic risk (see figures 6.2-

6.4). It should be noted, moreover, that these interactive effects persist independent of a host of confounding factors, including low birth weight, low attachment security, low household income, low maternal education, female-headed household, and postpartum depression.

As depicted in figure 6.2, among all subjects in the same-sex twin sample, those who had the highest level of genetic risk exhibited the best behavior when exposed to a very long duration of breastfeeding (e.g., >= 12 months), yet exhibited the worst behavior when they breastfed for less than 6 months. To be precise, the predicted externalizing behavior score for individuals who possess the highest degree of genetic risk, but were breastfed for at least 12 months, was 1.50229

(approximately the 14th percentile) after adjusting for covariates. Conversely, the predicted externalizing behavior score for individuals who possess the highest degree of genetic risk, but were breastfed for less than 6 months, was 3.15006 (approximately the 93rd percentile). Thus, the results suggest that breastfeeding duration is most relevant to the externalizing behavior of individuals with the highest level of genetic risk. A longer duration of breastfeeding, however, did not confer additional protection against behavioral problems during kindergarten among those with low genetic risk, despite increasing externalizing behavior somewhat among individuals with the lowest level of genetic risk (see figure 6.2).

Figure 6.3 provides a graphical display of the interaction between the dichotomous breastfeeding measure (with the 6-month threshold) and the genetic risk measure. Specifically, the figure depicts the interaction by identifying two groups distinguished by the 6-month

128 breastfeeding threshold, placing them on the x-axis (in descending order), fixing covariates to their means, and then plotting externalizing behavior scores on the y-axis on the basis of genetic risk scores within each of the breastfeeding groups. As depicted in the figure, among subjects who are at the lowest level of genetic risk, those who breastfed for 6 months or more are predicted to exhibit somewhat similar levels of externalizing behavior as those who breastfed for less than 6 months, net of covariates [predicted score is 1.87864 and 1.71148 respectively, or approximately the 37th and the 26th percentile]. In other words, differences in the duration of breastfeeding across the 6-month threshold appear to have only small effects on externalizing behavior at the lowest level of genetic risk, with subjects who breastfed for longer durations displaying slightly higher levels of externalizing behavior. Still, regardless of breastfeeding history, subjects with a lowest degree of genetic risk fall well below the median level of externalizing behavior for the sample.

In contrast, among subjects who possess the highest level of genetic risk, those who breastfed for less than 6 months score substantially higher on kindergarten externalizing behavior than those who breastfed for 6 months or more. After adjusting for covariates, the average externalizing behavior score for subjects who possess the highest level of genetic risk, yet breastfed for 6 months or more, is 2.4145, whereas the average externalizing behavior score for subjects who possess a high level of genetic risk and breastfed for less than 6 months is 3.15006.

This difference amounts to approximately 24 percentile points, as a score of 3.15006 falls approximately at the 94th percentile of externalizing behavior scores, while a score of 2.4145 falls at approximately the 70th percentile of externalizing behavior scores. Thus, while individuals with the highest level of genetic risk score well above the median regardless of

129 breastfeeding duration, such individuals are predicted to score abnormally high on externalizing behavior when they were also breastfed for fairly short durations (< 6 months).

Finally, figure 6.4 graphs the predicted externalizing behavior score at each of the four levels of genetic risk after splitting the sample into two groups: subjects who were breastfed exclusively for at least 6 months and subjects who were not breastfed exclusively for at least 6 months. The figure indicates that each unit increase in genetic risk results in a greater degree of externalizing behavior problems, but only for subjects who were not exclusively breastfed for 6 months or more (which constitutes the vast majority of the sample). However, among subjects who were exclusively breastfed for 6 months or more, additional units of genetic risk actually correspond to a lower degree of externalizing behavior. The clear conclusion gleaned from figure 6.4, therefore, is that exclusive breastfeeding strongly protects against externalizing behavioral problems, but only among individuals with a high degree of genetic risk.

To illustrate, the predicted externalizing behavior score among subjects who were breastfed exclusively for 6 months or more decreases by approximately 1.49 units when moving from the lowest to the highest level of genetic risk, suggesting that the group with the highest level of genetic risk benefit the most from long-term, exclusive breastfeeding. In contrast, the predicted externalizing behavior score among subjects who were not breastfed exclusively for at least 6 months increases by approximately 1.37 units when moving from the lowest to the highest level of genetic risk, suggesting that the group with the highest level of genetic risk is most adversely affected by a short duration of breastfeeding. Upon examining the predicted externalizing behavior scores of individuals with the highest degree of genetic risk, such scores vary drastically contingent on whether long-term, exclusive breastfeeding occurred. To be precise, subjects with the highest level of genetic risk, who were nevertheless breastfed

130 exclusively for at least 6 months, are predicted to have an externalizing behavior score of

.831467, which approximates the lowest score possible on the externalizing scale. On the other hand, subjects with the highest level of genetic risk who failed to breastfed exclusively for at least 6 months are predicted to have an externalizing behavior score of 3.1046 (or approximately the 92nd percentile).

In sum, both the table and the figures pertaining to the interactions between the breastfeeding measures and genetic risk (i.e., table 6.11 and figures 6.2-6.4) reveal an overarching pattern: a shorter duration of breastfeeding, regardless of operationalization, constitutes a noteworthy risk factor for externalizing behavior among individuals who possess a high degree of genetic risk. However, among individuals a lower degree of genetic risk, a shorter duration of breastfeeding does not appear to increase the externalizing behavior of offspring (and may even result in relatively low levels of externalizing behavior). A related, particularly striking finding is that the behavior of subjects who possess a high degree of genetic risk is most responsive to long-term, exclusive breastfeeding (i.e., they benefit the most from long-term, exclusive breastfeeding, yet are most adversely affected when it does not occur).

Apart from examining genetic risk as a moderator of the link between breastfeeding practices and offspring externalizing behavior, I also explore whether subjects who engaged in little to no breastfeeding as an infant are especially likely to exhibit high levels of externalizing behavior if they also experienced low attachment security during toddlerhood (see Table 6.12).

The models in table 6.12 are organized in a similar manner as the models in table 6.11 – that is, each model interacts the moderator (i.e., low attachment security) with the three breastfeeding measures (i.e., Short Duration of Breastfeeding, Breastfed Less than 6 Months, and Not

Exclusively Breastfed). As illustrated in table 6.12, low attachment security fails to significantly

131 moderate the relationship between breastfeeding practices and offspring externalizing behavior, regardless of how breastfeeding practices are measured. The implication of this finding is that the significant, direct effects of breastfeeding practices on externalizing behavior detected in prior analyses are not stronger (or weaker) for subjects who evinced low attachment security during toddlerhood. Thus, the link between breastfeeding practices and externalizing behavior does not appear to be significantly different across levels of attachment security.

Finally, low birth weight is examined as a potential moderator of the relationship between a short duration of breastfeeding and externalizing behavior during kindergarten (see table 6.13).

Prior evidence has suggested that the cognitive and physical development of low birth weight infants may be especially sensitive to breastfeeding initiation and duration, and that breastfeeding may help them to “catch up” developmentally to normal birth weight children

(Horwood et al., 2001; Lucas et al., 1992; Vohr et al., 2007). It is possible, therefore, that low birth weight infants may also exhibit relatively poor behavioral outcomes when breastfeeding is minimal or nonexistent.

I explore this possibility in models 1-3 of table 6.13. The results of these models suggest that, when employing the continuous measure of breastfeeding, a shorter duration of breastfeeding does not significantly interact with low birth weight to predict kindergarten externalizing behavior. Put differently, earlier weaning does not increase externalizing behavior to a greater degree for low birth weight children. Nevertheless, upon examining threshold effects in model 2 of table 6.13, the results indicate that breastfeeding for less than 6 months is particularly likely to amplify childhood externalizing behavior for subjects who were born low birth weight. Again, this interaction only emerged when breastfeeding was dichotomized at the

6-month threshold (see figure 6.5). Importantly, exclusivity of breastfeeding was irrelevant to

132 the interaction between breastfeeding for less than 6 months and low birth weight status, as evidenced in model 3. It is reaching or surpassing the 6-month mark, therefore, but not necessarily doing so exclusively, that appears to exert an especially protective influence on the externalizing behavior of low birth weight infants.

The results displayed in model 2 of table 6.13 are illustrated in figure 6.5. In the figure, subjects are grouped by breastfeeding threshold on the x-axis, and within each of these groups, subjects externalizing behavior scores are plotted based on their birth weight status (i.e., normal or low), with covariates being held at their mean. As depicted in the figure, subjects who are born low birth weight, yet breastfeed for at least 6 months, score significantly lower on the externalizing behavior scale than subjects who are born normal birth weight and were breastfed for at least 6 months (2.11943 v. 2.23032; approximately an 8 percentile point difference).

Conversely, individuals who are born low birth weight and breastfed for less than 6 months score significantly higher on the externalizing behavior scale than normal birth weight individuals who were breastfed for less than 6 months (2.17078 v. 2.03426; approximately an 8 percentile point difference).

While the effect of failing to reach the breastfeeding threshold on externalizing behavior is distinct across groups distinguished by birth weight status, it should be noted that this interaction effect is not as substantive as the interaction effects detected between breastfeeding practices and genetic risk. Specifically, regardless of which breastfeeding-birth weight group is examined, each group’s predicted externalizing behavior score is either close to the median score or slightly above it. Conversely, subjects with both a high level of genetic risk and a short duration of breastfeeding are predicted to score in the top 6-8% of the externalizing distribution

(depending on how breastfeeding is measured). In other words, although the interactions

133 between the breastfeeding threshold, genetic risk, and low birth weight are all statistically significant, figures 6.2-6.4 effectively illustrate that, in terms of the predicted externalizing behavior score, subjects who possess a high level of genetic risk and minimal exposure to breastfeeding appear to incur a much more pronounced risk of a seriously problematic degree of externalizing behavior (i.e., 92nd - 94th percentile). Comparatively, low birth weight subjects who fail to cross the 6-month threshold of breastfeeding are predicted to score moderately high on externalizing behavior (59th percentile), but not extremely high. Thus, while both interactions are statistically significant and informative, the GxE between breastfeeding duration and genetic risk appears to be especially noteworthy, as figures 6.2-6.4 illustrate.

6.2 Results: Low Diet Quality Analyses

The low diet quality analyses employ each of the analytical techniques described in chapter 5 (METHODS) in an effort to answer the following four research questions:

5. In what ways do children with especially poor dietary habits differ from other children?

6. Does a low quality diet during preschool significantly increase the risk of externalizing

behavior problems during kindergarten, independent of familial and genetic influences?

7. Is the relationship between a low quality diet during preschool and externalizing behavior

during kindergarten robust to indicators of low attachment security during toddlerhood

and externalizing behavior during preschool?

8. Is the influence of a low quality diet on kindergarten externalizing behavior moderated by

genetic risk and/or low attachment security?

134 6.2.1 Research Question 5

The fifth research question asks: In what ways do children with especially poor dietary habits differ from other children?

As mentioned in chapter 5 (METHODS), table 6.14 displays the descriptive statistics of all variables included in the low diet quality analyses for the same-sex twin sample. The univariate results reveal that the average low diet quality score in the sample is 11.86, which implies that, in this sample, diets skew slightly to the healthier side of the distribution, since the range of observed values if 2-29. When specific components of the diet are examined, descriptive statistics indicate that low vegetable consumption and low fruit consumption are more common among the same-sex sample, whereas high soda consumption and high fast food consumption are somewhat less common. The average scores on a number of control variables that were repeated at wave 3 have shifted since wave 1. For example, relative to wave 1, scores on low household income, maternal depression, low maternal education, and female-headed household at wave 3 have all decreased slightly, suggesting that, on average, factors related to the well-being and financial stability of the parents improved somewhat during the 4-year interim.

Additional variables that were either unavailable or inapplicable at wave 1 were also added to some of the low diet quality analyses, including parental withdrawal, corporal punishment, infrequent family meals, and no family food rules. Univariate statistics pertaining to these variables reveal, for example, that 40% of parents surveyed reported spanking their child at least once during the week prior to the interview. Nineteen percent of parents, moreover, reported having no family rules or routines concerning the foods their children eat. The average score on the infrequent family meals variable is 1.38, indicating that the average number of times

135 that families in the sample were eating dinner together at wave 3 was approximately 5-6 times a week. Finally, the most common parental withdrawal experience is one in which parents feel mild to moderate levels of withdrawal from their child, with few participants scoring high on the variable.

A measure of parent-reported externalizing behavior of offspring at wave 3 was also included in a number of the low diet quality analyses. The univariate statistics pertaining to externalizing behavior during preschool and kindergarten corroborate findings within the child development literature (see Thomas & Pope, 2012). That is, children, on average, exhibited a greater degree of externalizing behavior during preschool relative to kindergarten (2.42 v. 2.15).

These differences support the notion that it is more developmentally uncommon to exhibit very high levels of externalizing behavior during kindergarten. As was the case in the breastfeeding analyses, the sample is exactly 50% male and 42% nonwhite, with an average age of 69.46 months (or approximately 5 years and 9 months) at waves 4/5.

Table 6.15 contains the bivariate correlations between all variables used in any of the low diet quality analyses. As would be predicted, a number of the components of the low diet quality measure are significantly and positively correlated, and all dietary components are highly correlated with the low diet quality index. For example, salty snack, soda and sweets consumption are significantly and positively correlated, as are low fruit and low vegetable consumption. Perhaps most interesting, however, is that low fruit and low vegetable consumption are both either uncorrelated or negatively correlated with all other components of the diet employed in the study. Low vegetable consumption does, however, appear to be positively correlated with other child characteristics, such as sex (male = 1) and low attachment security. Thus, it seems that children who eat fewer fruits and/or vegetables are not necessarily

136 eating more “junk food”, even though the distinct types of junk food do tend to cluster together within individuals. Thus, while the low diet quality measure represents the cumulative dietary risk of each child, distinct types of risky eating behaviors do not always cluster together within the same individuals. These patterns in the data provide the rationale for separate examinations of each of the dietary components as predictors of externalizing behavior (in addition to the cumulative risk index), in order to determine which dietary components (if any) are more or less relevant to externalizing behavior. At the bivariate level, it seems that greater consumption of

“junk food” (e.g., sweets, soda) during preschool is positively correlated with externalizing behavior during kindergarten.

Apart from the components of diet during preschool, several other covariates were significantly correlated. For example, in this sample, nonwhite children are more frequently found in high risk family environments, at least as it pertains to low maternal education, low household income, maternal depression, and female-headed household. Nonwhite children are also more likely to consume a higher amount of salty snacks and soda relative to white children.

Also, children who are exposed to harsher parenting techniques (i.e., corporal punishment) are also more likely to engage in poorer dietary practices, especially when it comes to salty snack and soda consumption. Finally, as would be expected, children with a low quality diet are a) more likely to come from low-income, low-education households and b) less likely to eat dinner regularly with the family or be subject to any food-related rules or routines.

Table 6.16 displays the results of several difference of means t tests which compare child, maternal, and household covariates across the 90th percentile of low diet quality. Thus, in this portion of the analysis, children who scored in the top decile of the low diet quality measure are distinguished from those who did not, in order to determine the extent to which children with

137 especially poor dietary practices are significantly different from other children on key covariates.20 To be precise, the table illustrates 1) the average score on each covariate for each of the two groups distinguished by the 90th percentile of the low diet quality measure, 2) the difference between the two means on each covariate, 3) the t-value that corresponds to the difference of means and 4) the statistical significance of the difference of means as denoted by the obtained p-value. The tests yielded a number of important findings.

First, the results reveal that nonwhite children were more commonly found in the top decile of low diet quality. Specifically, while 52% of the children in the top decile of low diet quality were nonwhite, only 40% of individuals who scored below the 90th percentile of low diet quality were nonwhite (t = 2.47, p = .01), a difference of 12 percentage points. Furthermore, children who were spanked by their parents at least once during the week prior to the interview were significantly more common among the worst eaters (52% v. 37%, t = 3.13, p = .00). On a related note, children with the worst nutritional habits during preschool also scored significantly higher on concurrent externalizing behavior (t = 4.16, p = .00).

The difference of means t tests also revealed that several characteristics of the mother and the household are related to children’s eating habits during preschool. For instance, the top decile on low diet quality scored, on average, nearly 2 units higher on low household income (t =

3.28, p = .00) and over a full unit higher on low maternal education (t = 3.70, p = .00). This suggests that, on average, children with the worst eating habits will reside in households with anywhere from $10,000 to $20,000 less in annual income, relative to other children.

20 In the case of maternal/household covariates, the sample used to determine significance was the sample of households, not individual children, as there were two children per household in the same-sex twin sample. To determine group classification based on low diet quality, one child was randomly chosen from each household to create the household sample upon which these t tests are based. However, in the case of child covariates, which were measured separately for each child within each household, the sample of individual children (not households) was used to determine significance.

138 Additionally, the results imply that children with the worst eating habits tend to have mothers who were unable to attend college, whereas the remaining children tend to have mothers who obtained at least some college education. Children who scored in the top decile on low diet quality were more likely to live in households with no family food rules/routines as well as infrequent family meals. To be precise, while nearly 37% of poor diet households (i.e., top decile on low diet quality) reported having no family food rules or routines, only 16% of the remaining 90% of households reported the same, a nearly 20 percentage-point difference (t =

3.67, p = .00). Moreover, on average, children with the poorest diets eat meals with their families approximately 5 days a week, whereas the remaining children eat meals with their families approximately 6 days a week (t = 2,23, p = .03). In sum, dietary practices of children during preschool appear to be related to a number of child, maternal, and household covariates, including child race, corporal punishment, externalizing behavior, low household income, low maternal education, infrequent family meals, and no family food rules.

Tables 6.17 through 6.22 display the results of several differences of means t tests which distinguish groups by the 90th percentile of poor diet quality on each of the six components of diet quality examined in the current study: low vegetable consumption (Table 6.17), low fruit consumption (Table 6.18), high fast food consumption (Table 6.19), high sweets consumption

(Table 6.20), high salty snack consumption (Table 6.21), and high soda consumption (Table

6.22). I opted to examine each dietary component separately, in addition to an examination of the cumulative index, due to the correlational patterns between the diet components displayed in table 6.15. That is, in light of the results depicted in table 6.15, there is reason to believe that low vegetable and low fruit consumption may be differentially related to key covariates

139 compared to the “junk food” components. Separate difference of means t tests for each dietary component will help to better elucidate these differences.

First, table 6.17 suggests that males are significantly more likely to be found among those who ate few, if any, vegetables. Specifically, while 59% of subjects who scored in the top decile of low vegetable consumption were males, only 47% of the remaining subjects (the bottom 90%) were male (t = 2.94, p = .00). Children who ate the fewest vegetables also tended to score higher on concurrent externalizing behavior (t = 2.16, p = .03). Perhaps most interestingly, is that children who consumed the fewest vegetables scored, on average, .11 units higher on low attachment security, suggesting that low vegetable consumption is associated with lower levels of mother-child attachment (t = 3.50, p = .00). Although .11 units may seem miniscule, in reality it corresponds to a 12 percentile-point difference on low attachment security between groups designated by the 90th percentile of low vegetable consumption (67th percentile v. 55th percentile). Importantly, this difference of means did not emerge as significant when groups were distinguished on the cumulative measure of low diet quality, and therefore appears to be unique to specific components of the diet.

In comparison to child covariates, there is little in the way of household and/or maternal covariates that differentially characterizes the groups distinguished by low vegetable consumption. The only significant difference pertains to the parental withdrawal variable, which suggests that parental withdrawal is significantly more common in households where children eat very few (if any) vegetables (t = 2.11, p = .04). It should be noted, however, that the groups designated by vegetable consumption showed marginally significant differences in terms of the proportion of subjects living in female-headed households. Specifically, children with the highest scores on low vegetable consumption were more frequently found in single mother

140 households, relative to other children (21% v. 13%; t = 1.84; p = .07). However, these differences did not quite attain statistical significance.

Table 6.18 examines covariates across the 90th percentile of low fruit consumption.

While some patterns are similar to those displayed in table 6.18, there are some noticeable differences. For example, while males seem to be more common among children who rarely, if ever, eat vegetables, they are not overrepresented among children who rarely, if ever, eat fruit.

Nevertheless, externalizing behavior and low attachment security scores are significantly higher among children who eat little to no fruit (relative to children who more commonly eat fruit) (t =

3.02, p = .00; t = 2.31, p = .02). Table 6.18 also reveals that children who rarely eat fruit tend to live in households characterized by lower incomes and less education, which, interestingly enough, was not the case with children who rarely eat vegetables (t = 1.94, p = .05; t = 2.10, p =

.04). Furthermore, children who rarely consumed fruit were significantly more likely to live in a female-headed household than children who consumed fruit more often (26% v. 13%; t = 2.81, p

= .01). Interestingly, neither food rules nor the frequency of family meals were associated with membership in the top decile of low fruit or low vegetable consumption.

Table 6.19 displays the results of several difference of means t tests across the 90th percentile of high fast food consumption. In terms of child-specific covariates, only race and corporal punishment emerged as significant. In particular, 54% of those who consumed fast food very frequently were nonwhite, whereas only 40% of those who consumed fast food less frequently were nonwhite (t = 2.75, p = .01). Moreover, corporal punishment was more commonly administered to children who ate a high amount of fast food (48% v. 37%, t = 2.07, p

= .04). Similar to table 6.18, children who consume fast food very frequently tend to reside in households typified by a greater degree of socioeconomic and educational disadvantage. Low

141 maternal education scores, in particular, were significantly higher in households where children ate fast food frequently (t = 3.10, p = .00). Finally, having no family food rules or routines was a more common practice in households where children frequently ate fast food (32% v. 17%, t =

2.45, p = .01).

Table 6.20 displays the results of several difference of means t tests across the 90th percentile of high sweets consumption. Overall, neither child-specific nor household covariates are significantly different across the 90th percentile of sweets consumption, suggesting that high sweets consumption may be a component of diet that is more difficult to predict. Nevertheless, externalizing behavior appeared to be slightly more common among subjects who ate sweets most frequently (t = 1.97, p = .05). Still, the usual demographic, child, and family factors that are associated with other components of the diet do not appear to distinguish the children who eat sweets very frequently from the children who do not.

Table 6.21 displays the results of several difference of means t tests across the 90th percentile of high salty snack consumption. In contrast to the results depicted in table 6.20, several child and household characteristics distinguish the two groups. Children in the high salty snack group (relative to those who score below the 90th percentile) are more commonly nonwhite

(t = 3.40, p = .00) and younger (t = -2.59, p = .01). Moreover, children in the high salty snack group are relatively more likely to live in households characterized by less income (t = 4.27, p =

.00), less maternal education (t = 2.86, p = .00), and no family food rules (t = 3.82, p = .00). As was the case in table 6.17, annual income differences between the groups range, on average, between $10,000 and $20,000. Maternal education differences between the two groups, however, are very small in real units (both groups typically attain some college education).

142 Finally, table 6.22 displays the results of several difference of means t tests across the

90th percentile of high soda consumption. The results reveal that children who are nonwhite, are subjected to corporal punishment in the home, and are less securely attached to their caregiver are more likely to find themselves consuming a high amount of soda (t = 3.25, p =.00; t = 3.63, p

= .00; t = 2.30, p = .02). In terms of household and maternal covariates, both household income and maternal education are significantly lower in homes where child subjects drink soda very frequently, relative to homes where child subjects drink soda less frequently (t = 4.20, p = .00; t

= 4.64, p =.00). Conversely, children growing up in a single-mother household are more commonly found in the high soda consumption group (21% v. 13%; t = 2.14, p = .03). Finally, of the children who fall within the top decile of soda consumption, 34% are not subjected to any family rules or routines concerning food, while only 15% of the remaining children live in households with no family food rules (a 20 percentage-point difference; t = 4.76, p = .00).

To summarize, the covariates that appear to be the most relevant to whether children find themselves among the worst eaters are (from most to least relevant) 1) low household income

(sig. in 5 out of 7 models), 2) low maternal education (sig. in 5 out of 7 models), 3) nonwhite

(sig. in 4 out of 7 models), 4) no family food rules (sig. in 4 out of 7 models), 5) corporal punishment (sig. in 3 out of 7 models), and 6) low attachment security (sig. in 3 out of 7 models).

Some child and household covariates, however, were not very useful in identifying children with the worst eating habits, including the age of the child, their genetic risk for externalizing behavior, maternal depression, and parental withdrawal. Furthermore, some of the most relevant covariates in identifying patterns of “junk food” consumption were not found to be relevant covariates in identifying patterns of low vegetable consumption (e.g., low household income, low maternal education). Conversely, low attachment security was associated with especially

143 low consumption of healthy foods (e.g., fruits and vegetables), but not especially high consumption of junk foods (with the exception of soda). Ultimately, the pattern of results detected in this series of t tests establishes the non-randomness of child dietary patterns and highlights the need for a research design that can effectively rule out alternative explanations of the potential associations between dietary factors during preschool and externalizing behavior during kindergarten (e.g., within-family, genetically informative designs).

6.2.2 Research Question 6

The sixth research question asks: Does a low quality diet during preschool significantly increase the risk of externalizing behavior problems during kindergarten, independent of familial and genetic influences?

Table 6.23 displays the descriptive statistics pertaining to the DF analyses examining the influence of preschool dietary factors on externalizing behavior. The table reveals that high soda and fast food consumption are, on average, less common than low fruit and vegetable consumption. The difference scores also indicate substantial variation between twins within a twin pair on all dimensions of the diet measured (63% discordance on the comprehensive diet measure). For example, in the case of both low vegetable and fruit consumption, some twin pairs emerged in which one twin ate vegetables and/or fruit four or more times a day, but their cotwin never ate vegetables and/or fruit (this would correspond to a difference score of 6 or -6).

Similar variation in dietary habits between twins within a twin pair was also detected across the other dimensions of the diet. Of the dietary measures, however, fast food consumption showed the smallest degree of variation between twins within a twin pair (range: -4 to 4). As was the case in the breastfeeding analyses, low attachment security was highly variable among many twins in the same household. Externalizing behavior at wave 3 also showed a notable degree of

144 variation within twin pairs, with the largest twin difference being 2.57 units (where the maximum possible difference would be 4).

I now turn to the DF portion of the low diet quality analyses. Table 6.24 contains the results of eight models exploring the influence of the shared environment, heritability, and preschool dietary factors on externalizing behavior during kindergarten. Model 1 displays the results of the baseline model with no nonshared environmental effects specified. Model 2 includes the composite measure of low diet quality during childhood as a nonshared environment in the DF equation. Finally, models 3 through 8 examine whether specific components of childhood nutrition are more likely to lead to externalizing behavioral problems than others. The results of model 1 reiterate the results displayed previously: approximately 83% of the variance in externalizing behavioral problems during kindergarten can be attributed to genetic factors, with the remaining proportion of the variance being attributable to nonshared environmental factors and error. Model 2 expands model 1 by including the composite measure of low diet quality as a nonshared source of variance. The results of model 2 suggest that a low quality diet during the preschool years significantly increases the degree of externalizing behavior problems during kindergarten, even after taking genes and the shared environment into account. More precisely, the results suggest that, within twin pairs, the twin with lower overall diet quality tends to exhibit a significantly greater degree of externalizing behavior during kindergarten (again, twins evinced diet discordance in about 63% of cases).

When the six nutritional components that comprise the low diet quality measure are examined individually, only some of them emerge as having an independent influence on externalizing behavior during kindergarten. For example, the results from model 4 suggest that twins who consume fewer servings of fruit at age 4, relative to their cotwin, engage in a

145 significantly greater degree of externalizing behavioral problems during kindergarten, independent of genetic and shared environmental effects. Similar findings were also obtained when high fast food consumption, high sweets consumption, and high salty snack consumption were examined (models 5, 6, and 7). However, in the remaining models examining each specific dietary component, twin differences in eating patterns at wave 3 did not significantly impact their relative level of externalizing behavior at wave 4/5. To be precise, although the composite measure of diet at wave 3 significantly influenced externalizing behavior at wave 4/5, nutritional components related to low vegetable consumption and high soda consumption did not significantly impact externalizing behavior once the influence of genes and the shared environment were taken into account.

In sum, the results suggest that an early-childhood diet low in fruit, yet high in sweets, fast foods and salty snacks, is especially likely to heighten conduct problems during kindergarten, even when genetic factors and shared environmental factors are accounted for in the models. However, twins who consume a relatively high amount of soda and/or a relatively low amount of vegetables do not appear to exhibit significantly greater levels of externalizing behavioral problems relative to their cotwin, despite the fact that the composite measure of low diet quality at age 4 does have a significant impact on externalizing behavior during kindergarten.

6.2.3 Research Question 7

The seventh research question asks: Is the relationship between a low quality diet during preschool and externalizing behavior during kindergarten robust to indicators of low attachment security during toddlerhood and externalizing behavior during preschool?

146 The analysis proceeds with an examination of the possibility that low attachment security might explain the link between dietary factors and externalizing behavior. Research to date has revealed that a) a low level of attachment security is associated with poorer dietary habits

(Bozorgi et al., 2014; Faber & Dubé, 2015; Goossens et al., 2011; Lu et al., 2013) and b) low attachment security increases the likelihood of conduct problems (Dubois-Comtois et al., 2013;

Groh et al., 2012; Kochanska & Kim, 2013; O’Connor, et al., 2012). In light of this body of research, it is possible that twin differences in eating behaviors may not predict their relative level of externalizing behavior once differences in their security of attachment are taken into account.

To examine this possibility, I first conduct a DF analysis regressing twin scores on each of the dietary components on twin differences in low attachment security. The results of these analyses are displayed in table 6.25. Specifically, model 1 examines the influence of low attachment security on low diet quality, independent of genetic and shared environmental factors, whereas the remaining models (2-7) examine the influence of low attachment security on each of the dietary components. Interestingly, regardless of which dietary component is examined, the analysis revealed two key findings.

First, variation in eating behaviors is significantly influenced by both shared and nonshared environmental process, but does not appear to be significantly influenced by genetic factors. Specifically, anywhere from 54-89% of the variance in the dietary factors is attributable to the shared environment. Interestingly, the shared environment explained a greater portion of the variance in salty snack and fast food consumption (.87, .89) than low vegetable and low fruit consumption (.54, .58). Due to the nonsignificance of the genetic term in each case, the nonshared environment, in conjunction with error, is therefore predicted to explain anywhere

147 from 11-46% of the variance in dietary practices during preschool. Second, after accounting for shared environmental influences, twin differences in low attachment security do not predict their relative risk of engaging in poor eating behaviors.21 These results contradict much of the prior literature in this area of research (Bost et al., 2014; Bozorgi et al., 2014; Lu et al., 2013). It is important to note, however, that the current analysis represents a more rigorous, internally valid test of the research question, and that prior associations between attachment security and diet quality may, in reality, have been spurious due to important shared environmental and/or genetic factors that were not taken into account.

Table 6.26 displays the results of a DF analysis examining whether the effects of dietary factors on externalizing behavior persist independent of low attachment security. Each of the eight models reveals that twin differences in low attachment security significantly predict differences in their level of externalizing behavior during kindergarten. However, differences in low attachment security do not appear to explain the relationship between twin differences in diet and their relative level of externalizing behavior. Several components of diet during early childhood, as well as the composite diet measure, significantly impact externalizing behavior, independent of the level of attachment security.22 Thus, it seems that both diet and attachment have significant, independent effects on externalizing behavior, reaffirming that the quality of the diet during early childhood is not merely a proxy for the quality of the child-caregiver relationship.

An additional concern that emerges when examining the influence of preschool dietary factors on kindergarten externalizing behavior is whether the effects of dietary factors on

21 Sensitivity analyses revealed that is independent (b = -.50; p = .03) was the only attachment hotspot that predicted low diet quality. 22 Upon examining individual hotspots, the results remained the same. In short, regardless of which attachment hotspot was examined, attachment items failed to confound the relationship between low diet quality and externalizing behavior.

148 externalizing behavior are robust to prior levels of externalizing behavior. Put differently, is the link between preschool diet and kindergarten conduct problems merely an indicator of stability in conduct problems from preschool to kindergarten? A potentially important reason that twins with unhealthier eating patterns have a greater degree of subsequent behavioral problems

(relative to their cotwin) may be because their poor eating is merely a manifestation of a more challenging disposition. In short, eating discordance between twins may suggest a household dynamic in which parents are trying to provide a healthy, balanced diet to both twins, but the twin with the more challenging, noncompliant personality and/or behavior merely refuses to consume the healthy options that are offered to him. In this way, poor eating habits may be a reflection of difficult temperament and/or behavior, which was already reported by parents at wave 3 (externalizing behavior wave 3). Thus, it is possible to model the influence of preschool diet on kindergarten externalizing behavior, independent of preexisting behavioral problems, to further explore the robustness of dietary influences on externalizing behavior.

The set of DF models testing this possibility is presented in table 6.27. Again, model 1 displays the results of the baseline model with no nonshared environmental effects specified.

Model 2 includes the composite measure of low diet quality as a nonshared environment in the

DF equation. Finally, models 3 through 8 examine whether specific components of childhood nutrition are more likely to lead to externalizing behavioral problems than others. Importantly, twin differences in prior levels of externalizing behavior are also modeled as a nonshared source of variance. As in prior analyses, model 1 reveals that about 83% of the variance in externalizing behavior is attributable to genetic factors, with the remaining variance being explained by the nonshared environment and error.

149 Model 2 expands model 1 by including the composite measure of low diet quality as a nonshared source of variance. Because a challenging temperament might underpin poor nutritional habits as well as poor behavior, we included a measure of existing behavioral problems at age 4. As would be expected, preexisting differences in externalizing behavior between twins are highly predictive of subsequent twin differences in externalizing behavior problems (b = .29-.31, p = .00). Notwithstanding, the results of model 2 suggest that a low quality diet during the preschool years significantly increases the degree of externalizing behavior problems during kindergarten, even after taking genes, the shared environment, and prior externalizing behavior into account. The influence of low diet quality during preschool on externalizing behavior during kindergarten is therefore robust to the inclusion of the parent-rated measure of child externalizing behavior at age 4, which buttresses the notion that the relationship detected between poor eating habits and poor behavior is not merely tapping stability in poor behavior.

Models 3 through 8 examine whether the individual components of the diet during preschool still have an impact on kindergarten externalizing behavior, independent of preexisting behavioral problems. Interestingly, a number of significant effects emerge, although they are not entirely consistent with the results presented in table 5. Specifically, the results suggest that twins who consume fewer servings of vegetables at age 4, relative to their cotwin, displayed a significantly greater degree of externalizing behavioral problems during kindergarten. Similar findings were also obtained when low fruit consumption and high sweets consumption were examined (models 4 and 6). However, in the remaining models examining each specific component of diet, twin differences in eating patterns at wave 3 did not significantly impact their relative level of externalizing behavior at wave 4/5. To be precise, although the composite

150 measure of diet at wave 3 significantly influenced externalizing behavior at wave 4/5, nutritional components related to fast food consumption, salty snack consumption, and soda consumption did not significantly impact externalizing behavior once genes, the shared environment, and preexisting externalizing behavior were taken into account.

To summarize, the overall pattern of results suggests that poor dietary practices during preschool increase the likelihood of externalizing behavioral problems during kindergarten. In general, these findings are robust to indicators of low attachment security and preexisting externalizing behavior problems. The statistical significance of specific dietary dimensions, however, varied slightly across models. For example, low vegetable consumption only predicted kindergarten externalizing behavior once preexisting behavioral problems were taken into account, whereas high fast food consumption only predicted externalizing behavior when preexisting behavioral problems were not taken into account. Nonetheless, the current analyses significantly contribute to the body of literature linking early childhood diet to externalizing behavior (Oh et al., 2013; Park et al., 2012; Woo et al., 2014), as they are the first to do so using a genetically informative sibling design.

6.2.4 Research Question 8

The eighth research question asks: Is the influence of a low quality diet on kindergarten externalizing behavior moderated by genetic risk and/or low attachment security?

Table 6.28 displays the models that include interactions between the early childhood nutrition measures and the measure of genetic risk. Importantly, all of the models in these tables include all of covariates listed in table 6.14 (the descriptives table for the low diet quality analyses). Overall, there is little evidence to suggest that the influence of nutritional factors at the age of 4 on kindergarten externalizing behavior significantly varies as a function of genetic

151 risk. Of the seven models displayed across table 6.28, only one interaction term was statistically significant. Specifically, model 4 of table 6.28 suggests that the influence of frequent fast food consumption on subsequent externalizing behavior in enhanced under conditions of high genetic risk (see figure 6.6). That is, the results of the interactive model suggest that fast food consumption may be an especially noteworthy environmental risk factor for individuals who possess a high level of genetic risk for externalizing behavior.

As depicted in figure 6.6, focal children who possess the highest degree of genetic risk and consumed fast food at least once a day during the week prior to the wave 3 interview were found to have a predicted externalizing behavior score of 3.19095 (or approximately the 94th percentile), after adjusting for covariates. Conversely, subjects with a high degree of genetic risk who did not eat fast food during the week prior to the interview were predicted to score, on average, 2.67836 (or approximately the 81st percentile) on the externalizing behavior measure, after adjusting for covariates. Thus, while a high degree of genetic risk clearly increases the average predicted value of externalizing behavior during kindergarten, the genetic risk is exacerbated (by 13 percentile points, to be precise) when it occurs in conjunction with very frequent fast food consumption (see figure 6.6). Furthermore, externalizing behavior scores are predicted to increase by approximately .74 units from the lowest to the highest level of genetic risk among subjects who consumed no fast food during the week prior to the interview.

Conversely, externalizing behavior scores are predicted to increase by nearly twice that amount

(approximately 1.39 units) from the lowest to the highest level of genetic risk among subjects who consumed fast food every day during the week prior to the interview. Importantly, no other component of diet appears to have this amplifying effect on the relationship between genetic risk and externalizing behavior.

152 Table 6.29 displays the results of the interactions between each of the measures of early childhood nutrition (six component measures and one composite measure) and low attachment security during toddlerhood. In the DF models, several dietary components during preschool as well as low attachment security during toddlerhood predicted externalizing behavior during kindergarten. A related question that has yet to be examined is whether these independent effects might be amplified or attenuated in the presence of the other. For example, poor nutrition during preschool might have little bearing on kindergarten externalizing behavior if a child is securely attached to his/her mother, or conversely, poor nutrition might be especially impactful when attachment security is low (a positive interaction). These possibilities were empirically examined in the current study and the results are presented in table 6.29.

Importantly, all of the models in these tables include all of same covariates employed in table 6.29 (e.g., corporal punishment, no family food rules, parental withdrawal, etc.).

Regardless of the specific dietary component examined, no statistically significant interactions were detected between poor diet and low attachment security, suggesting that low attachment security neither exacerbates nor diminishes the effect of low diet quality on externalizing behavior during kindergarten. The clear implication is that the significant direct effects of diet detected in previous DF analyses are not significantly different across levels of attachment security.

In sum, there is minimal evidence to suggest that the influence of a low quality diet on externalizing behavior is moderated by genetic risk. The only dietary component that significantly interacted with genetic risk was high fast food consumption. As the interaction was positive, the results suggest that children who eat fast food more frequently are especially likely to develop externalizing behavioral problems during kindergarten if they also possess a relatively

153 high degree of genetic risk. Additional models revealed no evidence of an interactive relationship between low attachment security and low diet quality in the prediction of kindergarten externalizing behavior. Thus, while both low attachment security and low diet quality increased externalizing behavioral scores in the DF models, product-term analyses suggest that these effects operate independently, rather than interactively.

6.3 Results: Nutritional Interplay Analyses

The nutritional interplay analyses employ each of the analytical techniques described in chapter 5 (METHODS) in an effort to answer the following three research questions:

9. Does a short duration of breastfeeding increase the likelihood of a low quality diet?

10. Is the relationship between short duration of breastfeeding and externalizing behavior

problems during kindergarten explained by a low quality diet during preschool?

11. Do a short duration of breastfeeding and a low quality diet during preschool interact to

predict kindergarten externalizing behavior?

6.3.1 Research Question 9

The ninth research question asks: Does a short duration of breastfeeding increase the likelihood of a low quality diet?

The final section of results examines the interplay between nutritional factors across infancy and early childhood in the prediction of externalizing behavior during kindergarten.

While the previous two sections of the current study have explored breastfeeding practices and preschool eating habits separately, a number of studies to date have suggested that it is worthwhile to explore the interconnectedness of nutritional factors across infancy and early childhood (Abraham et al., 2012; Grieger et al., 2011; Perrine et al., 2014; Scott et al., 2012).

154 Doing so can help to elucidate the ways in which breastfeeding practices and preschool eating behaviors might work together to influence the likelihood of externalizing behavioral problems during kindergarten.

Table 6.30 simply presents the univariate statistics of the variables pertaining to this section of results. Prior sections have provided these same statistics, but the interplay between these variables has not yet been explored. Table 6.31 displays several bivariate correlations between the breastfeeding measures and the early childhood nutrition measures. The pattern of results that are presented in the table suggest that breastfeeding practices and preschool eating behaviors are correlated. To illustrate, low diet quality and short duration of breastfeeding are significantly and positively correlated (.22, p = .00), as are low diet quality and breastfed less than 6 months (.20, p = .00). The table also reveals that virtually all components of the low diet quality measure (e.g., low vegetable consumption, low fruit consumption, etc.) are positively and significantly correlated with a shorter duration of breastfeeding. Two interesting exceptions, however, emerge. First, high sweets consumption does not appear to be significantly correlated with the duration of breastfeeding. Second, exclusivity of breastfeeding is either unrelated or very weakly related to measures of preschool nutrition. Thus, it seems that the duration of breastfeeding, rather than the exclusivity of breastfeeding, is most relevant in the prediction of subsequent child eating habits.

In order to explore these patterns further, I conducted a number of difference of means t tests. The results of these tests are displayed in table 6.32. Specifically, scores on each of the three breastfeeding measures are obtained across the 90th percentile of each of the preschool dietary factors (the composite measures and the six components). The results reveal a number of interesting findings.

155 First, the average difference in the duration of breastfeeding across the 90th percentile of low diet quality was 1.75 months. To be precise, subjects who engaged in especially poor eating behaviors during preschool, on average, engaged in nearly two fewer months of breastfeeding, relative to other subjects (t = 3.34, p = .00). Similar patterns were detected when examining the frequency with which subjects reached the 6-month breastfeeding threshold. A staggering 93% of subjects who scored within the top decile of low diet quality failed to reach the 6-month mark, compared to 78% of the remaining subjects (t = 3.12, p = .00). Interestingly, all subjects who failed to exclusively breastfeed for at least 6 months also scored in the top decile of low diet quality during preschool. Nevertheless, the incidence of nonexclusive breastfeeding among the worst eaters was only marginally significantly different from the incidence of nonexclusive breastfeeding among those with better diets (t = 1.93, p = .05).

Second, the average difference in the duration of breastfeeding across the 90th percentile of low vegetable consumption was 1.21 months. To be precise, subjects who ingested an especially low amount of vegetables during preschool, on average, engaged in 1.21 fewer months of breastfeeding, relative to other subjects (t = 2.85, p = .00). Similar patterns were detected when examining the frequency with which subjects reached the 6-month breastfeeding threshold.

Approximately 89% of subjects who scored within the top decile of low vegetable consumption failed to reach the 6-month mark, compared to 78% of the remaining subjects (t = 3.12, p = .00).

Interestingly, 99% of subjects who failed to exclusively breastfeed for at least 6 months also scored in the top decile of low vegetable consumption during preschool. The incidence of nonexclusive breastfeeding among this group, however, was not significantly different from the incidence of nonexclusive breastfeeding among those who consumed vegetables more frequently.

156 Third, the average difference in the duration of breastfeeding across the 90th percentile of low fruit consumption was 1.61 months. In other words, subjects who rarely, if ever, consumed fruit during preschool, on average, engaged in 1.61 fewer months of breastfeeding, relative to other subjects (t = 3.60, p = .00). Similar patterns were detected when examining the frequency with which subjects attained the 6-month breastfeeding threshold. About 91% of subjects who scored within the top decile of low fruit consumption failed to reach the 6-month mark, compared to 78% of the remaining subjects (t = 3.35, p = .00). There was, however, virtually no difference between subjects in the incidence of exclusive breastfeeding across the 90th percentile of low fruit consumption (.97 versus .96).

Fourth, the average difference in the duration of breastfeeding across the 90th percentile of high fast food consumption was 1.58 months. This implies that subjects who consumed fast food very often, on average, engaged in 1.58 fewer months of breastfeeding, relative to other subjects (t = 3.02, p = .00). Similar patterns were detected when examining the frequency with which subjects reached the 6-month breastfeeding threshold. About 90% of subjects who scored within the top decile of high fast food consumption failed to reach the 6-month mark, compared to 79% of the remaining subjects (t = 2.58, p = .01). Interestingly, all subjects who failed to exclusively breastfeed for at least 6 months also scored in the top decile of high fast food consumption during preschool. Nevertheless, the incidence of nonexclusive breastfeeding among this group was only marginally significantly different from the incidence of nonexclusive breastfeeding among those who consumed fast food less frequently during preschool (t = 1.95, p

= .05).

Fifth, the average difference in the duration of breastfeeding across the 90th percentile of high sweets consumption was 1.13 months. In other words, subjects who consumed sweets very

157 frequently during preschool, on average, engaged in 1.13 fewer months of breastfeeding, relative to other subjects (t = 2.14, p = .03). About 88% of subjects who scored within the top decile of high sweets consumption failed to breastfeed for at least 6 months, compared to 79% of the remaining subjects. Despite the 9 percentage point difference, the difference in means between the groups is not statistically significant (t = 1.90, p = .06). Interestingly, the incidence of breastfeeding exclusivity (or the failure to do so for at least 6 months) was not significantly different across groups designated by the 90th percentile of sweets consumption (.95 versus .96).

Sixth, the average difference in the duration of breastfeeding across the 90th percentile of high salty snack consumption was 1.30 months. Put differently, subjects who ingested a high amount of salty snacks, on average, engaged in 1.30 fewer months of breastfeeding, relative to other subjects (t = 2.14, p = .03). About 85% of subjects who scored within the top decile of high salty snack consumption failed to breastfeed for at least 6 months, compared to 78% of the remaining subjects (t = 2.02, p = .04). Interestingly, 98% of subjects who failed to exclusively breastfeed for at least 6 months also scored in the top decile of salty snack consumption during preschool. The incidence of nonexclusive breastfeeding among this group, however, was not significantly different from the incidence of nonexclusive breastfeeding among those who consumed salty snacks less frequently (.95).

Finally, the average difference in the duration of breastfeeding across the 90th percentile of high soda consumption was 2.01 months. On average, subjects who drank soda (and other sugary drinks) very frequently engaged in approximately two fewer months of breastfeeding, relative to other subjects (t = 5.20, p = .00). About 90% of subjects who scored within the top decile of high soda consumption failed to breastfeed for at least 6 months, compared to 77% of the remaining subjects (t = 3.62, p = .00). Importantly, 99% of subjects who failed to

158 exclusively breastfeed for at least 6 months also scored in the top decile of high soda consumption during preschool. The incidence of nonexclusive breastfeeding among those who consumed soda less frequently was slightly, but significantly, lower (.95).

To summarize, individuals who engaged in particularly poor eating practices during preschool tended to breastfeed for significantly shorter durations. With the exception of the breastfeeding exclusivity measure, this relationship was largely robust to the different dietary components and different measures of breastfeeding. Thus, at the bivariate level, there is substantial evidence to suggest that a relatively short duration of breastfeeding and a relatively poor diet during preschool go hand in hand, even though exclusivity of breastfeeding does not seem to play much of a role in this relationship.

Of course, the co-occurrence of a short breastfeeding duration and poor eating during early childhood might be explained by a number of environmental factors (e.g., parental education, socioeconomic status, parenting practices) as well as genetic factors and/or genetically influenced traits (e.g., poor temperament, fussiness, low self-control). In order to rule out the effects of the shared environment and genetic factors, it is necessary to move beyond bivariate analyses to within-family, genetically sensitive designs. Tables 6.33-6.35 examine each of the three breastfeeding measures as predictors of preschool dietary factors using DF analysis. The results reveal several key findings.

First, as noted in previous models, the results reveal that a significant portion of the variance in eating behaviors during preschool is due to the shared environment (54-89%) and that the remaining variance (11-46%) in eating behaviors during preschool is due to a combination of the nonshared environment and error. Second, and more importantly, the results displayed in table 6.33 reveal that twin differences in the duration of breastfeeding predict the

159 relative quality of their overall diet during preschool, with twins who breastfed for fewer months exhibiting poorer overall eating habits during preschool. Due to the modeling strategy, neither shared environmental nor genetic factors have the potential to confound this relationship, which increases the internal validity of the finding. When specific components of the diet are examined, twins who breastfeed for shorter durations consumed a greater amount of salty snacks at age 4 relative to their cotwin. Otherwise, other dietary components are not independently predicted by breastfeeding duration. Therefore, it seems that differences in breastfeeding duration lead to significant differences in broad dietary patterns by age 4, even though these patterns are generally not restricted to any particular component of the diet (except salty snack consumption).

Upon examining the dichotomous measures of breastfeeding, however, there is little evidence to suggest that twins who fail to reach or surpass the 6-month threshold, whether exclusively or nonexclusively, are at any greater risk of poor dietary habits than their cotwins who did reach or surpass the threshold. That is, models 1 through 7 in tables 6.34 and 6.35 find no significant impact of breastfeeding less than 6 months or not exclusively breastfeeding on the composite measure of low diet quality or any of the components of the composite measure.

Thus, the effects of breastfeeding practices on offspring diet during preschool appear to be limited to dose-response effects. These results suggest that additional months of breastfeeding can lead to an improvement in overall diet quality and a reduction in salty snack consumption, regardless of when the child is ultimately weaned and whether breastfeeding is exclusive.

160 6.3.2 Research Question 10

The tenth research question asks: Is the relationship between short duration of breastfeeding and externalizing behavior problems during kindergarten explained by a low quality diet during preschool?

Due to the significant association between breastfeeding practices and dietary patterns during preschool, as well as the influence of these nutritional factors on externalizing behavior, it is possible that low diet quality functions as a mediating mechanism that explains the relationship between a short duration of breastfeeding and externalizing behavior. Table 6.36 displays the results for several DF analyses examining this research question. The results of each model indicate that, when the breastfeeding threshold variable and each of the dietary variables are examined in conjunction, the pattern of results remains the same as in previous analyses.

Specifically, breastfeeding for less than 6 months remains a significant predictor of externalizing behavior, even after accounting for preschool diet quality. Furthermore, in addition to the low diet quality measure, a number of dietary factors during preschool continue to influence externalizing behavior during kindergarten, independent of whether the breastfeeding threshold was attained (e.g., low fruit, high fast food, high sweets, and high salty snack consumption).

The results of the DF analysis displayed in table 6.36, in conjunction with the results of

DF analyses presented previously, reaffirms that a) breastfeeding for less than 6 months predicts increased externalizing behavioral problems during kindergarten and b) this relationship is not explained by the mediators examined in the current study. Specifically, neither low diet quality during preschool (table 6.36) nor low attachment security during toddlerhood (table 6.9) explain this breastfeeding threshold effect. However, the body of results gleaned from DF analyses also suggests that, independent of direct threshold effects, there are indirect, dose-response effects of

161 breastfeeding on externalizing behavior. Specifically, results thus far have indicated that, independent of genetic and shared environmental factors, fewer months of breastfeeding, independent of any threshold, corresponds to lower attachment security (table 6.7) as well as lower diet quality (table 6.33), both of which, in turn, predict higher externalizing behavior scores (see table 6.8 and table 6.24). Figure 6.7 provides a visual illustration of this pattern of results using a DF analysis nested within a path model (χ2 = 13845.03, df = 17, p < .05; RMSEA

= 1.07, p < .05).23 The figure clearly depicts the interconnectedness of nutritional factors during infancy and early childhood, attachment security during toddlerhood, and conduct problems during kindergarten. Ultimately, the figure illustrates the relevance of both threshold and dose- response effects of breastfeeding practices on externalizing behavior, while highlighting the significant mediating role of low attachment security and low diet quality in the dose-response model.

6.3.2 Research Question 11

The eleventh and final research question asks: Do a short duration of breastfeeding and a low quality diet during preschool interact to predict kindergarten externalizing behavior?

The final research question examined in the current study pertains to whether nutritional factors across infancy and early childhood interact to predict conduct problems during kindergarten. Children’s behavioral profile may be differentially influenced by their exposure to breastfeeding, contingent on their diet during preschool. Children who did not have the opportunity to breastfed, or only did so for a very short duration, may have more to gain developmentally from a balanced diet. Conversely, these same children may also be less

23 Importantly, the nature of the DF modeling strategy does not permit excellent model fit. To enhance model fit would require fitting covariances that undermine the logic of the DF design. The only purpose in conducting a DF analysis nested within a path model in the current study is to estimate effects simultaneously that have already been estimated independently.

162 resilient to the effects of a poor diet on their development. These possibilities are examined in

18 product-term analyses, the results of which are summarized in table 6.37. The table displays the estimates garnered when externalizing behavior during kindergarten was regressed on the product of scores on each of the six dietary components and scores on each of the three breastfeeding measures.24 Overall, the analyses provide little support for the hypothesis that nutritional factors across infancy and early childhood interact to predict externalizing behavior during kindergarten. There are, however, a few exceptions.

First, product-term analysis revealed that the positive effect of low vegetable consumption on externalizing behavior only emerged for children who were breastfed for less than 6 months (see figure 6.8). To illustrate, children who were breastfed less than 6 months and who ate no vegetables the week prior to the wave 3 interview received a predicted externalizing behavior score of 2.19771 (approximately the 60th percentile). By comparison, children who were breastfed for 6 months or more, but who ate no vegetables the week prior to the wave 3 interview, received a substantially lower predicted externalizing behavior score of 1.75

(approximately the 26th percentile). Thus, a lower frequency of vegetable consumption during preschool appears to increase the externalizing behavior of children, but only when they are breastfed for less than 6 months. Children who breastfeed for 6 months or more, however, appear to be quite resilient to the risk of externalizing behavior incurred though the infrequent consumption of vegetables, despite scoring relatively high on externalizing behavior when they eat vegetables very frequently.

Second, an additional product-term analysis revealed that the effect of breastfeeding less than 6 months on externalizing behavior was contingent on the degree of salty snack

24 The same covariates that were used in the interactive analyses in section 6.1 (breastfeeding analyses) were included in each of the product-term analyses in this section (section 6.3), including the genetic risk variable.

163 consumption at age 4 (see figure 6.9). For instance, children who were breastfed less than 6 months and who ate salty snacks four or more times per day during the week prior to the wave 3 interview received a predicted externalizing behavior score of 2.37528 (approximately the 69th percentile). Conversely, children who were breastfed for 6 months or more, but who ate salty snacks four or more times per day during the week prior to the wave 3 interview, received a substantially lower predicted externalizing behavior score of 1.78571 (approximately the 32nd percentile). Thus, frequent salty snack consumption during preschool appears to increase externalizing behavior, but only for children who were breastfed for less than 6 months.

Breastfeeding for 6 months or more, however, appears to protect children who rarely eat vegetables and frequently eat salty snacks from developing externalizing behavior problems.

Finally, a significant, negative interaction was detected between the “not exclusively breastfed” measure and the high sweets consumption measure. That is, the results suggest that the risk of externalizing behavior incurred by the frequent consumption of sweets is significantly diminished for those who failed to engage in long-term, exclusively breastfeeding (see figure

6.10). In other words, whether sweets are eaten frequently at age 4 appears to have little impact on the externalizing behavior of children who did not engage in long-term, exclusive breastfeeding, despite having a sizable impact on the externalizing behavior of children who did engage in long-term, exclusive breastfeeding. To illustrate, children who participated in long- term, exclusive breastfeeding, and consumed no sweets during the week prior to the wave 3 interview, are predicted to have an externalizing behavior score of 1.47437 (about the 12th percentile). Alternatively, children who participated in long-term, exclusive breastfeeding, and consumed sweets daily during the week prior to the wave 3 interview, are predicted to have an externalizing behavior score of 2.4727 (about the 74th percentile). Among children who were not

164 exclusively breastfed, the predicted increase in externalizing behavior when moving from “no sweets consumed” to “sweets consumed daily” was less than 1/5 of the increase in externalizing behavior predicted for exclusively breastfed children (.17922 units versus .99833).

Ultimately, these results suggest a mixed pattern of findings. In general, however, there is little evidence to suggest that the various dietary components significant interact with breastfeeding practices to predict externalizing behavior during kindergarten. Furthermore, in the few instances where significant interactions emerge, they are not all in the same direction (2 are positive, 1 is negative). It should be noted, however, than the effects of 2 of the 6 indicators of poor diet quality on externalizing behavior were positive only for children who failed to reach the 6-month breastfeeding threshold. Thus, there is some evidence to suggest that the externalizing behavior of children who breastfed for less than 6 months is adversely impacted by low vegetable consumption and high salty snack consumption, even though this was not the case for children who breastfed for 6 months or more. Despite these findings, however, the overall dietary pattern across all six components does not appear to significantly interact with breastfeeding practices to increase or decrease scores on externalizing behavior (see table 6.38).

165 Table 6.1: Descriptive Statistics for the Same-Sex Twin Sample (Breastfeeding Analyses)

Variable Mean Standard Deviation Range

1. Externalizing Behavior (4/5) 2.15 .60 1-4.57

2. Short Duration of Breastfeeding 20.01 4.59 0-23

3. Breastfed Less than 6 Months .81 .39 0-1

4. Not Exclusively Breastfed .96 .18 0-1

5. Sex .50 .50 0-1

6. Race .42 .49 0-1 7. Age of Child (in months) 69.46 5.18 57.3-84.8 8. Low Household Income 5.39 3.45 0-12 9. Age of Mother (in years) 30.12 6.55 17-57 10. Postpartum Depression 1.46 .47 1-3.42 11. Low Maternal Education 5.44 1.95 1-9 12. Female-Headed Household .18 .38 0-1 13. Low Parental Involvement 2.33 .71 1.33-4.33 14. Low Attachment Security -.39 .37 -.98-.86 15. Low Birth Weight .57 .49 0-1 16. Genetic Risk .91 .62 0-3

166 Table 6.2: Bivariate Correlations (Breastfeeding Analyses)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

1 1 .12** .16** .07* .27** .07* .01 .19** -.09* .11** .16** .11** .03 .16** .01 .51**

2 1 .84** .43** .05 .16** -.02 .19** -.24** .14** .23** .20** .15** .16** .09** .10**

3 1 .40** .09** .16** -.02 .19** -.21** .14** .22** .17** .16** .15** .12** .13**

4 1 .07* .05 .06 .06 -.08* .02 .14** .09** .07 .04 .05 .07

5 1 -.05 .10** .01 -.05 -.08* .00 .02 .00 .10** -.03 .15**

6 1 -.02 .42** -.23** .24** .34** .34** .10** .16** .12** .05

7 1 .02 -.05 .02 -.02 -.05 -.06 .05 .01 .06

8 1 -.45** .37** .58** .41** .18** .13** .06 .05

9 1 -.23** -.40** .29** -.07 -.04 -.06 .05

10 1 .24** .23** .05 .06 .05 .10**

11 1 .23** .17** .13** .01 .02

12 1 .06 .12** .05 .04

13 1 .10** -.06 -.02

14 1 .03 .10**

15 1 .05

16 1

167 Table 6.3: Child and Maternal/Household Profiles across Breastfeeding Threshold (6 mos.)

< 6 mos. > 6 mos. Covariates (mean) (mean) Difference T-value P-value

Child-Specific Covariates

Sex (Male = 1) .52 .42 .10 2.48 .01

Race (Nonwhite = 1) .44 .33 .11 2.91 .00

Age of Child (in months) 69.45 69.67 -.22 -0.46 .65

Low Attachment Security -.36 -.49 .13 4.12 .00

Low Birth Weight .63 .50 .13 3.36 .00

Genetic Risk .95 .75 .20 3.61 .00

Maternal/Household Covariates

Low Household Income 5.59 4.59 1.00 2.49 .01

Age of Mother (in years) 29.38 32.14 -2.76 -3.95 .00

Postpartum Depression 1.49 1.35 .14 2.32 .02

Low Maternal Education 5.62 4.80 .82 3.65 .00

Female-Headed Household .21 .05 .16 3.46 .00

Low Parental Involvement 2.39 2.10 .29 3.54 .00

168 Table 6.4: Child and Maternal/Household Profiles by Exclusivity of Breastfeeding

Not Exclusive Exclusive > 6 mos. Covariates (mean) (mean) Difference T-value P-value

Child-Specific Covariates

Sex (Male = 1) .51 .33 .18 2.03 .04

Race (Nonwhite = 1) .42 .36 .06 .67 .50

Age of Child (in months) 69.53 67.93 1.60 1.61 .11

Low Attachment Security -.38 -.50 .12 1.75 .08

Low Birth Weight .60 .48 .12 1.42 .15

Genetic Risk .91 .68 .23 1.91 .06

Maternal/Household Covariates

Low Household Income 5.44 4.12 1.31 1.55 .12

Age of Mother (in years) 29.82 32.65 -2.83 -1.87 .06

Postpartum Depression 1.46 1.46 .00 .07 .95

Low Maternal Education 5.51 4.18 1.33 2.76 .01

Female-Headed Household .18 .00 .18 1.94 .05

Low Parental Involvement 2.34 2.16 .18 1.06 .29

169 Table 6.5: Descriptive Statistics of the Variables used in the Defries-Fulker Analyses (Breastfeeding Analyses)

Variable Mean Standard Deviation Range Outcome Variables Externalizing Behavior (W4/5) 2.15 .60 1-4.57 Low Attachment Security -.39 .37 -.98-.86 Covariates (Difference Scores) Short Duration of Breastfeeding 0 1.39 -20-20 Breastfed Less than 6 Months 0 .16 -1-1 Not Exclusively Breastfed 0 .11 -1-1 Low Attachment Security 0 .32 -1.52-1.52 Low Birth Weight 0 .48 -1-1

170 Table 6.6: DF analysis of the Shared Environment, Heritability, and Breastfeeding Practices as Predictors of Externalizing Behavior during Kindergarten

Externalizing Behavior (Wave 4/5) Model 1 Model 2 Model 3 Model 4 b SE b SE b SE b SE

Shared environment .00 .10 .00 .11 .00 .11 .00 .10

Heritability .83** .13 .84** .13 .86** .13 .79** .13

Nonshared Sources of Variance Short Duration of Breastfeeding .01 .01 - - - - Breastfed Less than 6 months - - .19** .07 - - Not Exclusively Breastfed - - - - .13 .10 N 760 740 740 724 R2 .25 .25 .25 .27 Notes: * p < .05; ** p < .01

171 Significant Twin Pair 1

Significant Twin Pair 2

Non-Sig. Twin Pair 3

Non-Sig. Twin Pair 4

Twin 1 0 3 6 9 12 15 18 21 24 Duration of Breastfeeding (in Months) Twin 2

Figure 6.1: An Illustration of the Breastfeeding Threshold Effect within Hypothetical Twin Pairs

172 Table 6.7: DF analysis of the Shared Environment, Heritability, and Breastfeeding Practices as Predictors of Low Attachment Security during Toddlerhood

Low Attachment Security (Wave 2) Model 1 Model 2 Model 3 Model 4 b SE b SE b SE b SE

Shared environment .52** .09 .56** .09 54** .09 .53** .10

Heritability .19 .13 .14 .13 .17 .13 .17 .13

Nonshared Sources of Variance Short Duration of Breastfeeding .03** .01 - - - - Breastfed Less than 6 months - - .01 .10 - - Not Exclusively Breastfed - - - - .15 .17 N 928 902 902 878 R2 .41 .43 .42 .41 Notes: * p < .05; ** p < .01

173 Table 6.8: DF analysis of the Shared Environment, Heritability, and Low Attachment Security as Predictors of Externalizing Behavior during Kindergarten

Externalizing Behavior (Wave 4/5) Model 1 Model 2 b SE b SE

Shared environment .00 .10 .02 .10

Heritability .83** .13 .78** .13

Nonshared Sources of Variance

Low Attachment Security - - .16** .06

N 760 720 R2 .25 .27

Notes: * p < .05; ** p < .01

174 Table 6.9: Does Low Attachment Security Explain the Link between Breastfeeding Practices and Externalizing Behavior?

Externalizing Behavior (Wave 4/5) Model 1 Model 2 Model 3 Model 4 b SE b SE b SE b SE

Shared environment .00 .10 .00 .10 .00 .11 .03 .11

Heritability .83** .13 .78** .13 .80** .13 .77** .13

Nonshared Sources of Variance Short Duration of Breastfeeding .00 .01 - - - - Breastfed Less than 6 months - - 23** .09 - - Not Exclusively Breastfed - - - - .08 .14 Low Attachment Security .17** .07 .17** .07 .18** .07 N 760 702 702 688 R2 .25 .27 .27 .28 Notes: * p < .05; ** p < .01

175 Table 6.10: Does Birth Weight Confound the Relationship between Breastfeeding Practices and Externalizing Behavior?25

Externalizing Behavior (Wave 4/5) Model 1 Model 2 Model 3 Model 4 b SE b SE b SE b SE

Shared environment .00 .10 .00 .11 .00 .11 .01 .10

Heritability .83** .13 .84** .13 .85** .13 .79** .13

Nonshared Sources of Variance

Short Duration of .01 .01 - - - - Breastfeeding

Breastfed Less than 6 - - .19** .07 - - Months

Not Exclusively - - - - .13 .10 Breastfed

Low Birth Weight .03 .04 .03 .04 .03 .04 N 760 740 740 724 R2 .25 .25 .25 .27 Notes: * p < .05; ** p < .01

25 A sensitivity analysis was conducted using a continuous measure of birth weight instead of the dichotomous low birth weight measure. The results were substantively identical.

176 Table 6.11: Is the Relationship between Breastfeeding Practices and Externalizing Behavior in Offspring Moderated by Genetic Risk?

Model 1 Model 2 Model 3 Beta/b Beta/b Beta/b Short Duration of .03 .02 - - Breastfeeding (.02) Breastfed Less than 6 - .04 .02 - Months (.02) Not Exclusively - - .05 .03 Breastfed (.02) Genetic Risk .40** .25 .41** .26 .43** .27 (.02) (.02) (.02) Low Attachment .06 .04 .06 .04 .08** .05 Security (.02) (.02) (.02) Low Birth Weight .01 .01 .01 .01 .01 .01 (.02) (.02) (.02) Sex (male = 1) .22** .27 .22** .27 .21** .25 (.04) (.04) (.04) Race (nonwhite = 1)26 -.02 -.03 -.03 -.03 -.02 -.03 (.05) (.05) (.05) Age of Child -.05 -.01 -.05 -.01 -.05 -.01 (.00) (.00) (.00) Low Household Income .12** .02 .12** .02 .12** .02 (.01) (.01) (.01) Age of Mother -.01 -.00 -.02 -.00 -.01 -.00 (.00) (.00) (.00) Low Maternal Education .08* .02 .07 .02 .08* .03 (.01) (.01) (.01) Postpartum Depression .02 (.02) .02 (.05) .00 (.00) (.05) (.05) (.05) Female-Headed .00 .00 .00 .00 .00 .00 Household (.06) (.06) (.06) Low Parental .02 .01 .02 .02 .02 .02 Involvement (.03) (.03) (.03) SDOB X Genetic Risk .12** .09 - - (.02) BLTSM X Genetic Risk - .11** .08 - (.03) NEB X Genetic Risk - - .11** .08 (.02) N 657 657 648 R2 .34 .34 .34 Notes: * p < .05; ** p < .01. SDOB = Short Duration of Breastfeeding. BLTSM = Breastfed Less than 6 Months. NEB = Not Exclusively Breastfed.

26 Alternative measures of Race/Ethnicity, including specific measures of black and Hispanic, did not alter the results of any of the product-term analyses in the current dissertation.

177 3.2

2.8

2.6

2.4 Breastfed: < 6 months 2.2 6-11 months >= 12 months 2 Externalizing Behavior (W4/5) Behavior Externalizing 1.8

1.6

1.4 0 1 2 3 Genetic Risk

Figure 6.2: The Predicted Externalizing Behavior Score (W4/5) By Level of Genetic Risk and Short Duration of Breastfeeding When Covariates Are at Their Mean

178 3.2

2.8

2.6 Genetic Risk: 2.4 Lowest Low 2.2 High

2 Highest Externalizing Behavior (W4/5)Behavior Externalizing 1.8

1.6

1.4 >= 6 months < 6 months Breastfeeding Threshold

Figure 6.3: The Predicted Externalizing Behavior Score (W4/5) By Breastfeeding Threshold (6 Months) and Level of Genetic Risk When Covariates Are at Their Mean

179

3.25

2.75

2.5

2.25 Genetic Risk:

Lowest 2 Low

1.75 High Highest 1.5 Externalizing Behavior (W4/5) (W4/5) Behavior Externalizing

1.25

0.75 Exclusive Not Exclusive Exclusivity of Breastfeeding (>= 6 months)

Figure 6.4: The Predicted Externalizing Behavior Score (W4/5) By Exclusivity of Breastfeeding (>= 6 Months) and Level of Genetic Risk When Covariates Are at Their Mean

180 Table 6.12: Is the Relationship between Breastfeeding Practices and Externalizing Behavior in Offspring Moderated by Low Attachment Security?

Model 1 Model 2 Model 3 Beta/b Beta/b Beta/b Short Duration of -.01 -.00 - - Breastfeeding (.03) Breastfed Less than 6 - .01 .01 - Months (.02) Not Exclusively - - .01 .01 Breastfed (.02) Genetic Risk .43** .27 .43** .27 .44** .27 (.02) (.02) (.02) Low Attachment .07* .04 .07* .04 .08* .05 Security (.02) (.02) (.02) Low Birth Weight .01 .00 .01 .00 .01 .01 (.02) (.02) (.02) Sex (male = 1) .22** .27 .22** .26 .21** .26 (.04) (.04) (.04) Race (nonwhite = 1) -.03 -.04 -.03 -.04 -.03 -.04 (.05) (.05) (.05) Age of Child -.05 -.01 -.05 -.01 -.05 -.01 (.00) (.00) (.00) Low Household Income .13** .02 .13** .02 .13* .02 (.01) (.01) (.01) Age of Mother -.02 -.00 -.02 -.00 -.02 -.00 (.00) (.00) (.00) Low Maternal Education .07 .02 .07 .02 .07 .02 (.01) (.01) (.01) Postpartum Depression .02 (.02) .02 (.02) .01 (.01) (.05) (.05) (.05) Female-Headed .00 .01 .00 .01 .00 .00 Household (.06) (.06) (.06) Low Parental .02 .02 .02 .02 .03 .03 Involvement (.03) (.03) (.03) SDOB X Low -.01 -.00 - - Attachment Security (.02) BLTSM X Low - .00 .00 - Attachment Security (.02) NEB X Low Attachment - - .05 .03 Security (.02) N 654 654 645 R2 .32 .32 .33 Notes: * p < .05; ** p < .01. SDOB = Short Duration of Breastfeeding. BLTSM = Breastfed Less than 6 Months. NEB = Not Exclusively Breastfed.

181 Table 6.13: Is the Relationship between Breastfeeding Practices and Externalizing Behavior in Offspring Moderated by Low Birth Weight?

Model 1 Model 2 Model 3 Beta/b Beta/b Beta/b Short Duration of .00 .00 - - Breastfeeding (.02) Breastfed Less than 6 - .02 .01 - Months (.02) Not Exclusively - - -.02 -.01 Breastfed (.02) Genetic Risk .43** .27 .44** .27 .44** .27 (.02) (.02) (.02) Low Attachment .07* .04 .07* .04 .08** .05 Security (.02) (.02) (.02) Low Birth Weight .01 .00 .00 .00 .01 .01 (.02) (.02) (.02) Sex (male = 1) .22** .27 .22** .26 .21** .25 (.04) (.04) (.04) Race (nonwhite = 1) -.03 -.04 -.03 -.04 -.03 -.03 (.05) (.05) (.05) Age of Child -.05 -.01 -.06 -.01 -.06 -.01 (.00) (.00) (.00) Low Household Income .13** .02 .13** .02 .12* .02 (.01) (.01) (.01) Age of Mother -.02 -.00 -.03 -.00 -.02 -.00 (.00) (.00) (.00) Low Maternal Education .07 .02 .06 .02 .08 .02 (.01) (.01) (.01) Postpartum Depression .02 (.02) .02 (.02) .01 (.01) (.05) (.05) (.05) Female-Headed .00 .00 .00 .00 .00 .00 Household (.06) (.06) (.06) Low Parental .02 .02 .02 .02 .03 .02 Involvement (.03) (.03) (.03) SDOB X Low Birth .05 .03 - - Weight (.02) BLTSM X Low Birth - .08** .05 - Weight (.02) NEB X Low Birth - - .06 .03 Weight (.02) N 657 657 648 R2 .32 .33 .33 Notes: * p < .05; ** p < .01. SDOB = Short Duration of Breastfeeding. BLTSM = Breastfed Less than 6 Months. NEB = Not Exclusively Breastfed.

182 2.25

2.2

) 2.15

2.1 Birth Weight:

Normal 2.05 Low

Externalizing Behavior (W4/5 Behavior Externalizing 2

1.95

1.9 >= 6 months < 6 months Breastfeeding Threshold

Figure 6.5: The Predicted Externalizing Behavior Score (W4/5) By Breastfeeding Threshold (6 Months) and Birth Weight Status When Covariates Are at Their Mean

183 Table 6.14: Descriptive Statistics for the Same-Sex Twin Sample (Low Diet Quality Analyses)

Variable Mean Standard Deviation Range

1. Externalizing Behavior (4/5) 2.15 .60 1-4.57

2. Low Diet Quality 11.86 3.87 2-29

3. Low Vegetable Consumption 2.99 1.44 0-6

4. Low Fruit Consumption 2.77 1.43 0-6

5. High Fast Food Consumption 1.00 1.01 0-6

6. High Sweets Consumption 2.14 1.28 0-6

7. High Salty Snack Consumption 1.64 1.18 0-6

8. High Soda Consumption 1.31 1.46 0-6

9. Sex .50 .50 17-57

10. Race .42 .49 0-1

11. Age of Child (in months) 69.46 5.18 57.3-84.8

12. Low Household Income (W3) 4.74 3.51 0-12

13. Maternal Depression (W3) 1.43 .49 1-4

14. Low Maternal Education (W3) 5.25 1.96 1-9

15. Female Headed Household (W3) .15 .35 0-1

16. Parental Withdrawal (W3) 2.08 .68 1-4

17. Corporal Punishment (W3) .40 .49 0-1

18. Infrequent Family Meals (W3) 1.38 1.84 0-7

19. No Family Food Rules (W3) .19 .39 0-1

20. Externalizing Behavior (W3) 2.42 .63 1-4.71

21. Low Attachment Security -.38 .37 -.98-.86

22. Genetic Risk .91 .62 0-3

184 Table 6.15: Bivariate Correlations (Low Diet Quality Analyses)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

1 1 .14** .04 .06 .05 .08* .08* .09** .27** .07* .01 .20** .21** .14** .06 .14**

2 1 .49** .50** .49** .45** .48** .57** .04 .04 .00 .11** .06 .12** .06 .05

3 1 .51** -.01 -.12** -.08* -.03 .10** -.03 .05 -.11** -.01 -.08* -.02 .05

4 1 -.02 -.13** -.07* .03 .05 -.01 .06 .08* .01 .06 .05 .01

5 1 .22** .22** .24** .01 .03 .04 .04 .05 .09** .04 .09*

6 1 .28** .19** .03 -.02 -.02 .01 .01 -.01 -.02 .01

7 1 .23** -.03 .07* -.10** .20** .03 .12** .05 .01

8 1 -.04 .07* -.04 .12** .09* .18** .09** .00

9 1 -.05 .10** .02 -.04 -.04 -.02 .03

10 1 -.02 .43** .12** .34** .29** .03

11 1 .-.01 -.03 -.04 -.07 .11**

12 1 .25** .58** .37** -.04

13 1 .19** .04 .25**

14 1 .19** -.10**

15 1 -.03

16 1

185 Table 6.15 Continued

17 18 19 20 21 22 1 .20** -.02 .03 .54** .16** .51** 2 .12** .17** .16** .10** .12** .03 3 -.04 .09* .00 .02 .12** .08* 4 .07 .06 .06 .03 .07 .03 5 .07 .15** .11** .06 .03 .07 6 .03 .10** .05 .07* -.02 -.06 7 .08* .07* .13** .02 .07 -.03 8 .16** .05 .15** .07* .07 -.01 9 .07* -.07* -.02 .13** .10** .15** 10 .02 .10** .20** .05 .16** .05 11 -.01 .01 .10** .06 .05 .06 12 .16** .03 .25** .14** .21** .08* 13 .12** .07 .06 .33** .13** .09* 14 .09* .08* .27** .11** .16** -.02 15 .09** .01 .06 .04 .13** .02 16 .10** -.02 .02 .20** -.03 .12** 17 1 .05 .06 .23** .04 .09** 18 1 .08* .01 .01 .03 19 1 .05 .10** -.02 20 1 .18** .26** 21 1 .10** 22 1

186 Table 6.16: Child and Maternal/Household Profiles across the 90th Percentile of Low Diet Quality

> 90th Pct < 90th Pct Covariates (mean) (mean) Difference T-value P-value

Child-Specific Covariates

Sex (Male = 1) .48 .49 .01 .30 .77

Race (Nonwhite = 1) .52 .40 .12 2.47 .01

Age of Child (in months) 69.77 69.43 .34 .55 .58

Corporal Punishment (W3) .52 .37 .15 3.13 .00

Externalizing Behavior (W3) 2.66 2.39 .27 4.16 .00

Low Attachment Security -.35 -.41 .06 1.35 .18

Genetic Risk .95 .91 .04 .60 .55

Maternal/Household Covariates

Low Household Income 6.36 4.54 1.82 3.28 .00

Maternal Depression 1.54 1.41 .13 1.70 .09

Low Maternal Education 6.20 5.13 1.07 3.70 .00

Female-Headed Household .22 .14 .08 1.43 .15

Parental Withdrawal (W3) 2.11 2.07 .04 .35 .73

Infrequent Family Meals (W3) 1.92 1.32 .60 2.23 .03

No Family Food Rules (W3) .37 .16 .21 3.67 .00

187 Table 6.17: Child and Maternal/Household Profiles across the 90th Percentile of Low Vegetable Consumption

> 90th Pct < 90th Pct Covariates (mean) (mean) Difference T-value P-value

Child-Specific Covariates

Sex (Male = 1) .59 .47 .12 2.94 .00

Race (Nonwhite = 1) .42 .41 .01 .29 .72

Age of Child (in months) 69.59 69.43 .16 .34 .74

Corporal Punishment (W3) .34 .39 -.05 -1.29 .20

Externalizing Behavior (W3) 2.51 2.40 .11 2.16 .03

Low Attachment Security -.31 -.42 .11 3.50 .00

Genetic Risk .96 .90 .06 1.00 .32

Maternal/Household Covariates

Low Household Income 4.93 4.70 .23 .51 .61

Maternal Depression 1.51 1.41 .10 1.57 .12

Low Maternal Education 5.43 5.22 .21 .88 .38

Female-Headed Household .21 .13 .08 1.84 .07

Parental Withdrawal (W3) 2.22 2.04 .18 2.11 .04

Infrequent Family Meals (W3) 1.60 1.33 .27 1.24 .22

No Family Food Rules (W3) .21 .18 .03 .63 .53

188 Table 6.18: Child and Maternal/Household Profiles across the 90th Percentile of Low Fruit Consumption

> 90th Pct < 90th Pct Covariates (mean) (mean) Difference T-value P-value

Child-Specific Covariates

Sex (Male = 1) .51 .49 .02 .47 .64

Race (Nonwhite = 1) .41 .41 .00 .04 .97

Age of Child (in months) 69.74 69.41 .33 .64 .52

Corporal Punishment (W3) .45 .37 .08 1.77 .08

Externalizing Behavior (W3) 2.57 2.39 .18 3.02 .00

Low Attachment Security -.33 -.41 .08 2.31 .02

Genetic Risk .96 .90 .06 .85 .39

Maternal/Household Covariates

Low Household Income 5.52 4.60 .92 1.94 .05

Maternal Depression 1.43 1.43 .00 .09 .93

Low Maternal Education 5.70 5.17 .53 2.10 .04

Female-Headed Household .26 .13 .13 2.81 .01

Parental Withdrawal (W3) 2.14 2.06 .08 .79 .43

Infrequent Family Meals (W3) 1.57 1.34 .23 .94 .25

No Family Food Rules (W3) .24 .18 .06 1.28 .20

189 Table 6.19: Child and Maternal/Household Profiles across the 90th Percentile of High Fast Food Consumption

> 90th Pct < 90th Pct Covariates (mean) (mean) Difference T-value P-value

Child-Specific Covariates

Sex (Male = 1) .48 .49 -.01 .21 .84

Race (Nonwhite = 1) .54 .40 .14 2.75 .01

Age of Child (in months) 70.15 69.37 .78 1.29 .20

Corporal Punishment (W3) .48 .37 .11 2.07 .04

Externalizing Behavior (W3) 2.50 2.41 .09 1.32 .19

Low Attachment Security -.40 -.40 .00 .02 .98

Genetic Risk 1.01 .90 .11 1.56 .12

Maternal/Household Covariates

Low Household Income 5.78 4.63 1.15 1.96 .05

Maternal Depression 1.47 1.43 .04 .53 .60

Low Maternal Education 6.09 5.16 .93 3.10 .00

Female-Headed Household .17 .15 .02 .45 .65

Parental Withdrawal (W3) 2.12 2.07 .05 .51 .61

Infrequent Family Meals (W3) 1.61 1.35 .26 .94 .35

No Family Food Rules (W3) .32 .17 .15 2.45 .01

190 Table 6.20: Child and Maternal/Household Profiles across the 90th Percentile of High Sweets Consumption

> 90th Pct < 90th Pct Covariates (mean) (mean) Difference T-value P-value

Child-Specific Covariates

Sex (Male = 1) .47 .49 -.02 -.52 .60

Race (Nonwhite = 1) .49 .40 .09 1.62 .11

Age of Child (in months) 69.50 69.46 .04 .06 .95

Corporal Punishment (W3) .45 .38 .07 1.38 .17

Externalizing Behavior (W3) 2.53 2.40 .13 1.97 .05

Low Attachment Security -.37 -.40 .03 .71 .48

Genetic Risk .90 .91 -.01 -.15 .88

Maternal/Household Covariates

Low Household Income 5.70 4.64 1.06 1.81 .07

Maternal Depression 1.46 1.43 .03 .36 .72

Low Maternal Education 5.5 5.23 .27 .89 .37

Female-Headed Household .13 .15 -.02 -.36 .72

Parental Withdrawal (W3) 2.20 2.06 .14 1.29 .20

Infrequent Family Meals (W3) 1.76 1.34 .42 1.48 .14

No Family Food Rules (W3) .22 .18 .04 .58 .56

191 Table 6.21: Child and Maternal/Household Profiles across the 90th Percentile of High Salty Snack Consumption

> 90th Pct < 90th Pct Covariates (mean) (mean) Difference T-value P-value

Child-Specific Covariates

Sex (Male = 1) .50 .49 .01 .38 .70

Race (Nonwhite = 1) .50 .38 .12 3.40 .00

Age of Child (in months) 68.65 69.75 -1.10 -2.59 .01

Corporal Punishment (W3) .42 .37 .05 1.22 .22

Externalizing Behavior (W3) 2.46 2.41 .05 1.09 .28

Low Attachment Security -.37 -.41 .04 1.20 .23

Genetic Risk .87 .93 -.06 -1.20 .23

Maternal/Household Covariates

Low Household Income 5.91 4.29 1.62 4.27 .00

Maternal Depression 1.40 1.44 -.04 -.76 .45

Low Maternal Education 5.69 5.09 .60 2.86 .00

Female-Headed Household .15 .15 .00 .24 .81

Parental Withdrawal (W3) 2.07 2.08 -.01 -.14 .89

Infrequent Family Meals (W3) 1.43 1.36 .07 .38 .71

No Family Food Rules (W3) .30 .15 .15 3.82 .00

192 Table 6.22: Child and Maternal/Household Profiles across the 90th Percentile of High Soda Consumption

> 90th Pct < 90th Pct Covariates (mean) (mean) Difference T-value P-value

Child-Specific Covariates

Sex (Male = 1) .50 .49 .01 .43 .67

Race (Nonwhite = 1) .51 .38 .13 3.25 .00

Age of Child (in months) 69.35 69.49 -.14 .33 .74

Corporal Punishment (W3) .49 .35 .14 3.63 .00

Externalizing Behavior (W3) 2.48 2.40 .08 1.61 .11

Low Attachment Security -.35 -.41 .06 2.30 .02

Genetic Risk .91 .91 .00 .07 .95

Maternal/Household Covariates

Low Household Income 6.09 4.36 1.73 4.20 .00

Maternal Depression 1.49 1.41 .08 1.48 .14

Low Maternal Education 6.00 5.02 .98 4.64 .00

Female-Headed Household .21 .13 .08 2.14 .03

Parental Withdrawal (W3) 2.06 2.08 -.02 -.29 .77

Infrequent Family Meals (W3) 1.36 1.38 -.02 -.13 .90

No Family Food Rules (W3) .34 .14 .20 4.76 .00

193 Table 6.23: Descriptive Statistics of the Variables used in the Defries-Fulker Analyses (Low Diet Quality Analyses)

Variable Mean Standard Deviation Range Outcome Variables Externalizing Behavior (W4/5) 2.15 .60 1-4.57 Low Diet Quality 11.86 3.87 2-29 Low Vegetable Consumption 2.99 1.44 0-6 Low Fruit Consumption 2.77 1.43 0-6 High Fast Food Consumption 1.00 1.01 0-6 High Sweets Consumption 2.14 1.28 0-6 High Salty Snack Consumption 1.64 1.18 0-6 High Soda Consumption 1.31 1.46 0-6 Predictors (Difference Scores) Low Diet Quality 0 2.64 -18-18 Low Vegetable Consumption 0 1.17 -6-6 Low Fruit Consumption 0 1.24 -6-6 High Fast Food Consumption 0 .68 -4-4 High Sweets Consumption 0 .86 -5-5 High Salty Snack Consumption 0 .98 -6-6 High Soda Consumption 0 1.04 -5-5 Externalizing Behavior (W3) 0 .62 -2.57-2.57 Low Attachment Security 0 .32 -1.52-1.52

194 Table 6.24: DF analysis of the Shared Environment, Heritability, and Preschool Dietary Factors as Predictors of Externalizing Behavior

Externalizing Behavior (Wave 4/5) Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 b SE b SE b SE b SE b SE b SE b SE b SE DF analysis components

Shared environment .00 .10 .00 .10 .00 .10 .00 .10 .00 .10 .00 .10 .00 .10 .00 .10

Heritability .83** .13 .80** .13 .83** .13 .84** .13 .83** .13 .81** .13 .82** .13 .83** .13

Nonshared Sources of Variance

Low Diet Quality .03** .01 ------

Low Vegetable Consumption - - .02 .02 ------

Low Fruit Consumption - - - - .03* .01 ------High Fast Food Consumption ------.06* .02 ------

High Sweets Consumption ------.10** .03 - - - - High Salty Snack Consumption ------.05* .02 - -

High Soda Consumption ------.01 .02 N 760 758 760 758 760 760 760 760 R2 .25 .27 .25 .25 .25 .26 .25 .25 Notes: * p < .05; ** p < .01

195 Table 6.25: DF analysis of the Shared Environment, Heritability, and Low Attachment Security as Predictors of Dietary Factors27

Low High Fast High Salty Low Diet Vegetable Low Fruit Food High Sweets Snack High Soda

Quality Consumption Consumption Consumption Consumption Consumption Consumption Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 b SE b SE b SE b SE b SE b SE b SE

DF Analysis Components

Shared environment .73** .09 .53** .10 .57** .11 .86** .14 .72** .09 .84** .12 .61** .10

Heritability .05 .12 .20 .14 .09 .17 .00 .21 .10 .13 .00 .17 .20 .15

Nonshared Source of Variance Low Attachment Security .23 .26 -.08 .10 .14 .13 -.03 .06 -.16 .09 .22 .12 .11 .10 N 858 862 760 862 860 862 862 R2 .58 .44 .39 .59 .61 .43 .55

Notes: * p < .05; ** p < .01

27 Baseline models, with each of the dietary factors at outcome variables, were also conducted. The proportion of the variance in the dietary components explained by the shared environment ranged from .54 to .89, depending on the dietary component examined. However, 72% of overall diet quality was explained by the shared environment, whereas none of the variance in diet quality was significantly influenced by genetic factors in these models. Ultimately, 28% of the variance in composite measure (low diet quality) can be attributed to a combination of the nonshared environment and error.

196 Table 6.26: Do the Effects of Dietary Factors on Externalizing Behavior Persist Independent of Low Attachment Security?

Externalizing Behavior (Wave 4/5) Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 b SE b SE b SE b SE b SE b SE b SE b SE DF analysis components

Shared environment .00 .10 .05 .10 .02 .10 .02 .10 .02 .10 .05 .10 .03 .10 .02 .10

Heritability .83** .13 .75** .13 .78** .13 .78** .13 .77** .13 .75** .13 .76** .13 .78** .13

Nonshared Sources of Variance

Low Diet Quality .03** .01 ------

Low Vegetable Consumption - - .03 .02 ------

Low Fruit Consumption - - - - .03* .01 ------High Fast Food Consumption ------.07** .03 ------

High Sweets Consumption ------.11** .03 - - - - High Salty Snack Consumption ------.04 .02 - -

High Soda Consumption ------.01 .02

Low Attachment Security .16** .06 .17** .06 .16** .06 .17** .06 .18** .06 .15* .06 .16** .06 N 760 718 720 718 720 720 720 720 R2 .25 .29 .27 .27 .28 .29 .27 .27 Notes: * p < .05; ** p < .01

197 Table 6.27: Do the Effects of Dietary Factors on Externalizing Behavior Persist Independent of Stability in Externalizing Behavior?

Externalizing Behavior (Wave 4/5) Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 b SE b SE b SE b SE b SE b SE b SE b SE DF analysis components

Shared environment .00 .10 .16 .10 .15 .10 .14 .10 .14 .10 .15 .10 .14 .10 .14 .10

Heritability .83** .13 .64** .13 .66** .13 .67** .13 .67** .13 .65** .13 .67** .13 .67** .13

Nonshared Sources of Variance

Low Diet Quality .03** .01 ------

Low Vegetable Consumption - - .04** .02 ------

Low Fruit Consumption - - - - .04** .01 ------High Fast Food Consumption ------.03 .02 ------

High Sweets Consumption ------.07** .03 - - - - High Salty Snack Consumption ------.03 .02 - -

High Soda Consumption ------.01 .02

Externalizing Behavior (W3) .30** .03 .31** .03 .31** .03 .30** .03 .29** .03 30** .03 30** .03 N 760 758 760 758 760 760 760 760 R2 .25 .37 .36 .36 .35 .36 .35 .35 Notes: * p < .05; ** p < .01

198 Table 6.28: Is the Relationship between Dietary Factors and Externalizing Behavior Moderated by Genetic Risk?

Externalizing Behavior (W4/5) Model 1 Model 2 Model 3 Beta/b Beta/b Beta/b Low Diet Quality .03 .02 - - (.02) Low Vegetable - .02 .01 - Consumption (.02) Low Fruit - - .04 .02 Consumption (.02) Genetic Risk .35** .21 .35** .21 .35** .21 (.02) (.01) (.02) Low Attachment .02 .04 .03 .03 .03 .04 Security (.05) (.05) (.05) Sex (male = 1) .15** .18 .15** .18 .15** .18 (.04) (.04) (.04) Race (nonwhite = 1) .00 .00 -.01 -.01 .00 .00 (.04) (.04) (.04) Age of Child -.05 -.01 -.05 -.01 -.06 -.01 (.00) (.00) (.00) Low Household Income .07 .01 .08* .01 .08 .01 (.01) (.01) (.01) Maternal Depression .00 .00 .00 .00 .00 .00 (.04) (.04) (.04) Low Maternal Education .06* .02 .07* .02 .06* .02 (.01) (.01) (.01) Female-Headed -.02 -.04 -.03 -.04 -.02 -.04 Household (.06) (.06) (.06) Parental Withdrawal .01 .01 .01 .01 .01 .01 (.03) (.03) (.03) Corporal Punishment .04 .05 .04 .06 .04 .05 (.04) (.03) (.04) Infrequent Family Meals -.03 -.01 -.02 -.01 -.02 -.01 (.01) (.01) (.01) No Family Food Rules -.01 -.02 .00 .00 -.01 -.01 (.05) (.05) (.05) Externalizing Behavior .40** .39 .40** .39 .40** .39 (W3) (.03) (.03) (.03) LDQ X Genetic Risk .03 .02 - - (.02) LVC X Genetic Risk - .02 .01 - (.02) LFC X Genetic Risk - - -.02 -.01 (.02) N 676 677 676 R2 .49 .49 .49 Notes: *p < .05; ** p < .01. LDQ = Low Diet Quality. LVC = Low Vegetable Consumption. LFC = Low Fruit Consumption

199 Table 6.28 (Continued)

Externalizing Behavior (W4/5) Model 4 Model 5 Model 6 Model 7 Beta/b Beta/b Beta/b Beta/b High Fast Food -.03 -.02 - - - Consumption (.01) High Sweets Consumption - .05 .03 - - (.02) High Salty Snack - - .05* .03 - Consumption (.02) High Soda Consumption - - - .05 .03 (.02) Genetic Risk .35** .21 .36** .22 .35** .21 .35** .21 (.02) (.02) (.02) (.02) Low Attachment Security .03 .05 .03 .05 .02 .04 .03 .05 (.05) (.05) (.05) (.05) Sex (male = 1) .16** .19 .15** .18 .15** .18 .15** .18 (.04) (.04) (.04) (.04) Race (nonwhite = 1) -.01 -.01 -.01 -.01 -.01 -.01 -.01 -.01 (.04) (.04) (.04) (.04) Age of Child -.05 -.01 -.05 -.01 -.05 -.01 -.05 -.01 (.00) (.00) (.00) (.00) Low Household Income .06 .01 .08 .01 .07 .01 .08 .01 (.01) (.01) (.01) (.01) Maternal Depression .01 .01 .00 .01 .01 .01 .00 .00 (.04) (.04) (.04) (.04) Low Maternal Education .07* .02 .07* .02 .06 .02 .06 .02 (.01) (.01) (.01) (.01) Female-Headed Household -.01 -.02 -.02 -.03 -.02 -.03 -.02 -.04 (.06) (.06) (.06) (.06) Parental Withdrawal .01 .01 .01 .01 .01 .01 .01 .01 (.03) (.03) (.03) (.03) Corporal Punishment .05 .06 .05 .06 .04 .05 .04 .05 (.03) (.03) (.04) (.03) Infrequent Family Meals -.02 -.01 -.02 -.01 -.02 -.01 -.02 -.01 (.01) (.01) (.01) (.01) No Family Food Rules .00 .00 .00 .00 -.01 -.01 -.01 -.01 (.05) (.05) (.05) (.05) Externalizing Behavior .40** .39 .40** .39 .41** .39 .40** .39 (W3) (.03) (.03) (.03) (.03) HFFC X Genetic Risk .08** .05 - - - (.02) HSWC X Genetic Risk - -.01 -.01 - - (.02) HSSC X Genetic Risk - - .03 .02 - (.02) HSOC X Genetic Risk - - - .02 .02 (.01) N 677 677 677 677 R2 .49 .49 .49 .49 Notes: *p < .05; ** p < .01. HFFC = Fast Food. HSWC = Sweets. HSSC = Salty Snacks. HSOC: Soda.

200 3.2

2.8

2.6 Fast Food Consumption: None 2.4 1-6 times per week 1 + times per day

2.2 Externalizing Behavior (W4/5) Behavior Externalizing

1.8 0 1 2 3 Genetic Risk

Figure 6.6: The Predicted Externalizing Behavior Score (W4/5) By Level of Genetic Risk and Fast Food Consumption When Covariates Are at Their Mean

201 Table 6.29: Is the Relationship between Dietary Factors and Externalizing Behavior Moderated by Low Attachment Security?

Externalizing Behavior (W4/5) Model 1 Model 2 Model 3 Beta/b Beta/b Beta/b Low Diet Quality .07* .04 - - (.01) Low Vegetable - .02 .01 - Consumption (.01) Low Fruit - - .04 .02 Consumption (.02) Genetic Risk .36** .22 .35** .21 .36** .21 (.02) (.02) (.02) Low Attachment .02 .01 .03 .02 .02 .01 Security (.02) (.02) (.02) Sex (male = 1) .15** .18 .15** .18 .15** .18 (.04) (.04) (.04) Race (nonwhite = 1) .00 .00 -.01 -.01 .00 .00 (.04) (.04) (.04) Age of Child -.05 -.01 -.05 -.01 -.05 -.01 (.00) (.00) (.00) Low Household Income .08 .01 .08* .01 .08 .01 (.01) (.01) (.01) Maternal Depression .00 .00 .00 .01 .00 .00 (.04) (.04) (.04) Low Maternal Education .06 .02 .07* .02 .06* .02 (.01) (.01) (.01) Female-Headed -.02 -.04 -.02 -.04 -.02 -.04 Household (.06) (.06) (.06) Parental Withdrawal .01 .01 .01 .01 .01 .01 (.03) (.03) (.03) Corporal Punishment .04 .05 .05 .06 .04 .05 (.03) (.03) (.04) Infrequent Family Meals -.03 -.01 -.02 -.01 -.02 -.01 (.01) (.01) (.01) No Family Food Rules -.01 -.02 .00 .00 .00 .00 (.05) (.05) (.05) Externalizing Behavior .40** .39 .40** .39 .41** .39 (W3) (.03) (.03) (.03) LDQ X Low Attachment .01 .01 - - Security (.02) LVC X Low Attachment - -.01 -.01 - Security (.02) LFC X Low Attachment - - .03 .02 Security (.02) N 676 677 676 R2 .49 .49 .49 Notes: *p < .05; ** p < .01. LDQ: Low Diet Quality. LVC = Low Vegetable Consumption. LFC: Low Fruit Consumption.

202 Table 6.29 (Continued)

Externalizing Behavior (W4/5) Model 4 Model 5 Model 6 Model 7 Beta/b Beta/b Beta/b Beta/b High Fast Food -.01 -.01 - - - Consumption (.02) High Sweets Consumption - .05 .03 - - (.02) High Salty Snack - - .05 .03 - Consumption (.02) High Soda Consumption - - - .05 .03 (.02) Genetic Risk .36** .22 .36** .22 .36** .22 .35** .21 (.02) (.02) (.02) (.02) Low Attachment Security .03 .02 .03 .02 .03 .02 .03 .02 (.02) (.02) (.02) (.02) Sex (male = 1) .15** .18 .15** .18 .15** .18 .15** .18 (.04) (.04) (.04) (.04) Race (nonwhite = 1) -.01 -.01 -.01 -.01 -.01 -.01 -.01 -.01 (.04) (.04) (.04) (.04) Age of Child -.05 -.01 -.05 -.01 -.05 -.01 -.05 -.01 (.00) (.00) (.00) (.00) Low Household Income .08 .01 .08 .01 .08 .01 .08* .01 (.01) (.01) (.01) (.01) Maternal Depression .00 .00 .00 .01 .01 .01 .00 .00 (.04) (.04) (.04) (.04) Low Maternal Education .07* .02 .07* .02 .07* .02 .06 .02 (.01) (.01) (.01) (.01) Female-Headed Household -.02 -.04 -.02 -.03 -.02 -.04 -.03 -.04 (.06) (.06) (.06) (.06) Parental Withdrawal .01 .01 .01 .01 .01 .01 .01 .01 (.03) (.03) (.03) (.03) Corporal Punishment .05 .06 .05 .06 .04 .05 .04 .05 (.04) (.03) (.04) (.03) Infrequent Family Meals -.02 -.01 -.02 -.01 -.03 -.01 -.02 -.01 (.01) (.01) (.01) (.01) No Family Food Rules .00 .00 -.01 -.01 -.01 -.01 -.01 -.01 (.05) (.05) (.05) (.05) Externalizing Behavior .40** .39 .40** .39 .40** .39 .40** .39 (W3) (.03) (.03) (.03) (.03) HFFC X Low Attachment .02 .01 - - - Security (.02) HSWC X Low Attachment - .02 .01 - - Security (.02) HSSC X Low Attachment - - -.05 -.03 - Security (.02) HSOC X Low Attachment - - - .00 .00 Security (.02) N 677 677 677 677 R2 .49 .49 .49 .49 Notes: *p < .05; ** p < .01. HFFC = Fast Food. HSWC = Sweets. HSSC: Salty Snacks. HSOC = Soda.

203 Table 6.30: Descriptive Statistics of the Nutritional Factors across Life Stages

Variable Mean Standard Deviation Range

1. Short Duration of Breastfeeding 20.01 4.59 0-23

2. Breastfed Less than 6 Months .81 .39 0-1

3. Not Exclusively Breastfed .96 .18 0-1

4. Low Diet Quality 11.86 3.87 2-29

5. Low Vegetable Consumption 2.99 1.44 0-6 6. Low Fruit Consumption 2.77 1.43 0-6 7. High Fast Food Consumption 1.00 1.01 0-6 8. High Sweets Consumption 2.14 1.28 0-6 9. High Salty Snack Consumption 1.64 1.18 0-6 10. High Soda Consumption 1.31 1.46 0-6

204 Table 6.31: Bivariate Correlations of Nutritional Factors across Life Stages

1 2 3 4 5 6 7 8 9 10 1 1 .82** .41** .22** .13** .11** .12** .03 .11** .15** 2 1 .39** .20** .13** .14** .10** .02 .06 .11** 3 1 .07* .07* .04 .07* -.03 .01 .05 4 1 .48** .50** .47** .47** .46** .58** 5 1 .51** -.01 -.12** -.11** -.03 6 1 -.01 -.12** -.09** .02 7 1 .20** .21** .22** 8 1 .26** .23** 9 1 .23** 10 1

205 Table 6.32: Breastfeeding Profiles across the 90th Percentile of Preschool Dietary Factors

> 90th Pct < 90th Pct Low Diet Quality (mean) (mean) Difference T-value P-value

Short Duration of Breastfeeding 22.47 20.72 1.75 3.34 .00

Breastfed Less than 6 Months .93 .78 .15 3.12 .00

Not Exclusively Breastfed 1.00 .96 .04 1.93 .05

Low Vegetable Consumption

Short Duration of Breastfeeding 21.90 20.69 1.21 2.85 .00

Breastfed Less than 6 Months .89 .78 .11 3.09 .00

Not Exclusively Breastfed .99 .96 .03 1.61 .11

Low Fruit Consumption

Short Duration of Breastfeeding 22.30 20.66 1.64 3.60 .00

Breastfed Less than 6 Months .91 .78 .13 3.35 .00

Not Exclusively Breastfed .97 .96 .01 .67 .50

High Fast Food Consumption

Short Duration of Breastfeeding 22.32 20.74 1.58 3.02 .00

Breastfed Less than 6 Months .90 .79 .11 2.58 .01

Not Exclusively Breastfed 1.00 .96 .04 1.95 .05

206 Table 6.32 (Continued)

> 90th Pct < 90th Pct High Sweets Consumption (mean) (mean) Difference T-value P-value

Short Duration of Breastfeeding 21.93 20.80 1.13 2.14 .03

Breastfed Less than 6 Months .88 .79 .09 1.90 .06

Not Exclusively Breastfed .95 .96 -.01 .52 .61

High Salty Snack Consumption

Short Duration of Breastfeeding 21.87 20.57 1.30 3.51 .00

Breastfed Less than 6 Months .85 .78 .07 2.02 .04

Not Exclusively Breastfed .98 .95 .03 1.65 .10

High Soda Consumption

Short Duration of Breastfeeding 22.47 20.46 2.01 5.20 .00

Breastfed Less than 6 Months .90 .77 .13 3.62 .00

Not Exclusively Breastfed .99 .95 .04 2.12 .03

207 Table 6.33: DF analysis of the Shared Environment, Heritability, and Short Duration of Breastfeeding as Predictors of Dietary Factors28

Low High Fast High Salty Low Diet Vegetable Low Fruit Food High Sweets Snack High Soda Quality Consumption Consumption Consumption Consumption Consumption Consumption Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 b SE b SE b SE b SE b SE b SE b SE DF Analysis Components

Shared environment .73** .08 .56** .10 .56** .11 .88** .13 .70** .08 .85** .12 .59** .10

Heritability .06 .11 .18 .13 .11 .16 .00 .21 .12 .13 .00 .17 .24 .14

Nonshared Source of Variance Short Duration of Breastfeeding .10** .03 .01 .02 .01 .02 -.01 .01 .00 .01 .07** .02 .00 .01 N 872 876 874 876 874 876 876 R2 .59 .46 .39 .60 .60 .44 .55 Notes: *p < .05; ** p < .01.

28 Baseline models, with each of the dietary factors at outcome variables, were also conducted. The proportion of the variance in the dietary components explained by the shared environment ranged from .54 to .89, depending on the dietary component examined. However, 72% of overall diet quality was explained by the shared environment, whereas none of the variance in diet quality was significantly influenced by genetic factors in these models. Ultimately, 28% of the variance in composite measure (low diet quality) can be attributed to a combination of the nonshared environment and error.

208 Table 6.34: DF analysis of the Shared Environment, Heritability, and Breastfed Less than 6 Months as Predictors of Dietary Factors

Low High Fast High Salty Low Diet Vegetable Low Fruit Food High Sweets Snack High Soda

Quality Consumption Consumption Consumption Consumption Consumption Consumption Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 b SE b SE b SE b SE b SE b SE b SE

DF Analysis Components

Shared

environment .73** .08 .56** .10 .55** .11 .88** .13 .70** .08 .84** .12 .59** .10

Heritability .07 .11 .18 .13 .12 .16 .00 .21 .12 .13 .00 .17 .24 .14

Nonshared Source of Variance Breastfed Less than 6 Months .48 .34 .15 .16 .00 .14 .00 .04 .00 .05 .31 .18 .00 .15 N 872 876 874 876 874 876 876 R2 .59 .46 .39 .60 .60 .43 .55 Notes: *p < .05; ** p < .01.

209 Table 6.35: DF analysis of the Shared Environment, Heritability, and Not Exclusively Breastfed as Predictors of Dietary Factors

Low High Fast High Salty Low Diet Vegetable Low Fruit Food High Sweets Snack High Soda

Quality Consumption Consumption Consumption Consumption Consumption Consumption Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 b SE b SE b SE b SE b SE b SE b SE

DF Analysis Components

Shared

environment .74** .08 .58** .10 .58** .11 .86** .13 .73** .08 .83** .12 .62** .10

Heritability .05 .11 .16 .13 .09 .16 .00 .21 .09 .13 .00 .17 .21 .14

Nonshared Source of Variance Not

Exclusively Breastfed .44 .76 .14 .18 .13 .18 .00 .05 .44 .29 -.14 .44 -.14 .12 N 848 852 850 852 850 852 852

R2 .60 .47 .41 .59 .62 .43 .57

Notes: *p < .05; ** p < .01.

210 Table 6.36: Is the Breastfeeding Threshold Effect on Externalizing Behavior Explained by Preschool Dietary Factors?

Externalizing Behavior (Wave 4/5) Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 b SE b SE b SE b SE b SE b SE b SE

DF analysis components

Shared environment .00 .10 .00 .11 .00 .11 .00 .11 .00 .10 .00 .11 .00 .11

Heritability .83** .13 .85** .13 .86** .13 .85** .13 .83** .13 .84** .13 .85** .13

Nonshared Sources of Variance

Breastfed Less than 6 Months .17* .08 .18** .07 .19** .07 .19** .07 .19** .07 .19** .07 .19** .07

Low Diet Quality .03** .01 ------

Low Vegetable Consumption - - .02 .02 ------

Low Fruit Consumption - - - - .04* .02 ------

High Fast Food Consumption ------.06* .03 ------

High Sweets Consumption ------.09** .03 - - - -

High Salty Snack Consumption ------.04* .02 - -

High Soda Consumption ------.01 .02 N 738 740 738 740 740 740 740 R2 .27 .25 .25 .25 .27 .25 .25

Notes: *p < .05; ** p < .01.

211

Breastfed Less than .19** Externalizing Behavior 6 Months (W4/5)

Low Attachment Security .03** .16**

Short Duration of Externalizing Behavior Breastfeeding (W4/5)

.13** Low Diet .03** Quality

Figure 6.7: An Illustration of the Direct and Indirect Effects of Breastfeeding Practices on Externalizing Behavior29

29 N = 700. (χ2 = 13845.03, df = 17, p < .05; RMSEA = 1.07, p < .05). Coefficients are unstandardized.

212 Table 6.37: Do Breastfeeding Practices and Preschool Dietary Factors Interact to Predict Externalizing Behavior in Offspring? Summary of Findings

Externalizing Behavior (W4/5) Low High High Vegetable Low Fruit Fast Food High Sweets Salty Snack High Soda Interaction Components Consumption Consumption Consumption Consumption Consumption Consumption

Short Duration of Breastfeeding .05 .03 -.01 -.00 -.07 -.06 -.05 -.03 .05 .05 -.04 -.03 (.02) (.02) (.03) (.02) (.03) (.03) Breastfed Less than 6 Months .06* .04 .00 .00 -.05 -.03 -.04 -.03 .08** .06 .00 .00 (.02) (.02) (.03) (.02) (.02) (.02) Not Exclusively Breastfed .00 .00 .01 .00 .07 .08 -.06** -.03 -.01 -.01 -.03 -.03 (.02) (.03) (.05) (.01) (.02) (.02) Notes: *p < .05; ** p < .01.

213 2.3

2.2

2.1

Breastfed: 2 < 6 Months >= 6 Months 1.9 Externalizing Behavior (W4/5) Behavior Externalizing

1.8

1.7 0 1 2 3 4 5 6 Low Vegetable Consumption

Figure 6.8: The Predicted Externalizing Behavior Score (W4/5) By Low Vegetable Consumption and Breastfeeding Threshold (6 Months) When Covariates Are at Their Mean

214 2.5

2.4

2.3

2.2 Breastfed: 2.1 < 6 Months >= 6 Months 2

Externalizing Behavior (W4/5) Behavior Externalizing 1.9

1.8

1.7 0 1 2 3 4 5 6 High Salty Snack Consumption

Figure 6.9: The Predicted Externalizing Behavior Score (W4/5) By High Salty Snack Consumption and Breastfeeding Threshold (6 Months) When Covariates Are at Their Mean

215 2.75

2.5

2.25

2 Breastfed:

Exclusively Nonexclusively 1.75

1.5

1.25

1 No Sweets Consumed Sweets Consumed Daily

Figure 6.10: The Predicted Externalizing Behavior Score (W4/5) By High Sweets Consumption and Exclusive Breastfeeding When Covariates Are at Their Mean

216 Table 6.38: Is the Relationship between Breastfeeding Practices and Externalizing Behavior in Offspring Moderated by Low Diet Quality?

Model 1 Model 2 Model 3 Beta/b Beta/b Beta/b Short Duration of -.03 -.02 - - Breastfeeding (.03) Breastfed Less than 6 - .00 .00 - Months (.02) Not Exclusively - - -.02 -.01 Breastfed (.02) Low Diet Quality

Genetic Risk .44** .28 .44** .27 .44** .28 (.02) (.02) (.02) Low Attachment .06 .03 .05 .03 .06* .04 Security (.02) (.02) (.02) Low Birth Weight .00 .00 .00 .00 .00 .00 (.02) (.02) (.02) Sex (male = 1) .22** .26 .22** .26 .21** .25 (.04) (.04) (.04) Race (nonwhite = 1) -.03 -.03 -.03 -.03 -.02 -.02 (.05) (.05) (.05) Age of Child -.05 -.01 -.05 -.01 -.05 -.01 (.00) (.00) (.00) Low Household Income .12** .02 .12** .02 .12* .02 (.01) (.01) (.01) Age of Mother -.02 -.00 -.01 -.00 -.02 -.00 (.00) (.00) (.00) Low Maternal Education .07 .02 .06 .02 .07 .02 (.01) (.01) (.01) Postpartum Depression .02 (.03) .02 (.02) .01 (.01) (.05) (.05) (.05) Female-Headed .01 .01 .00 .01 .00 .01 Household (.06) (.06) (.06) Low Parental .00 .00 .00 .00 .00 .00 Involvement (.03) (.03) (.03) SDOB X Low Diet -.02 -.01 - - Quality (.02) BLTSM X Low Diet - .02 .01 - Quality (.02) NEB X Low Diet - - -.04 -.03 Quality (.03) N 656 656 647 R2 .34 .33 .34 Notes: *p < .05; ** p < .01. SDOB = Short Duration of Breastfeeding. BLTSM = Breastfed Less than 6 Months. NEB = Not Exclusively Breastfed.

217 CHAPTER 7

DISCUSSION

This final chapter of the current dissertation has four objectives. The first objective is to provide a summary of the results of the present study and situate the results in the current literature. The second objective is to discuss the policy implications of the results of the current study. The third objective is to acknowledge some important limitations to the current study.

The fourth and final objective is to highlight important directions for future research on nutrition and antisocial outcomes.

7.1 Summary of Results

Chapter 6 included a host of tables and figures displaying the results of the present dissertation. Additionally, a large amount of text interpreting the results was provided in the previous chapter. In an effort to highlight some of broad patterns in the findings, in addition to the nuanced relationships across models, a summary of the results will now be presented.

In order to properly summarize the results, the eleven research questions asked in the current study will be presented and grouped by the nutritional factor being addressed (e.g., breastfeeding, diet quality, or both), just as they were presented in the results section. The first four research questions pertain to breastfeeding practices during infancy. The first research question asked whether, and in what ways, mother-child dyads who engage in breastfeeding for longer durations are different from mother-child dyads who engage in short-term breastfeeding

(or fail to breastfeed at all). The findings reveal that, in many ways, mother-infant pairs who practice long-term breastfeeding come from relatively advantageous backgrounds, particularly as it pertains to education and household income. Conversely, women who are running a household

218 without a partner are significantly less likely to breastfeed for at least 6 months. Women who breastfeed for short durations, furthermore, are more likely to exhibit a low level of parental involvement (e.g., lower frequency of singing, reading, and telling stories to the child). Upon examining child-specific characteristics, the results revealed that children who engage in long- term breastfeeding are more often female and white. On the other hand, children who have higher levels of genetic, perinatal, and socioemotional risk are less likely to engage in long-term breastfeeding. Furthermore, bivariate correlational analyses reveal that a shorter duration of breastfeeding is significantly and positively associated with externalizing behavior in this sample of children.

On the whole, mothers and children who fail to breastfeed for at least 6 months tend to score higher on several indicators of risk relative to mother-child dyads who attain or surpass the

6-month mark. Interestingly, the differential level of risk across groups is less robust when groups are distinguished by the exclusivity of breastfeeding. Still, mothers who engage in long- term, exclusive breastfeeding are, on average, significantly more educated and more likely to be living with a partner than mothers who fail to engage in long-term, exclusive breastfeeding.

Exclusivity of breastfeeding, however, was not significantly associated with group differences on any of the child-specific measures, with the exception of biological sex.

The second research question builds upon the first. It was expected that children with differential exposure to breastfeeding would differ not only in their conduct, but also in their exposure to a host of familial and individual risk factors that are also predictive of childhood conduct problems. The results pertaining to research question 1 largely confirmed these patterns, prompting the second research question, which asks whether a short duration of breastfeeding predicts externalizing behavior problems, net of shared environmental and genetic risk. DF

219 analysis of the data revealed that 83% of the variance in externalizing behavior is attributable to genetic factors, while none of the variance in externalizing behavior is attributable to the shared environment. Additionally, while a short duration of breastfeeding did not in and of itself have any significant, direct effects on externalizing behavior of offspring during kindergarten, failing to engage in breastfeeding for at least 6 months was found to significantly increase the level of externalizing behavior exhibited during kindergarten, net of shared environmental and genetic influences. Because the design compares twins within the same family, the detected associations cannot be spurious due to genes or shared environments that commonly contribute to a shorter duration of breastfeeding and an increased risk of offspring externalizing behavior (e.g., low maternal education, low household income, etc.).

Ultimately, these results provide mixed support for the relationship between length of breastfeeding and externalizing behavior, and suggest that the effect might be contingent on the exact stage at which the child is weaned. To illustrate, twin siblings who differ in their duration of breastfeeding are likely to show no significant differences in their level of externalizing behavior if they both breastfeed for relatively long durations (e.g., beyond 6 months).

Conversely, an extra month of two of breastfeeding appears to have little to no bearing on twin differences in externalizing behavior when both twins are weaned fairly soon after birth (e.g. 1 month v. 3 months). These results highlight not only the importance of a genetically informative, sibling comparison design to increase the validity of the results, but also the importance of testing for threshold effects in this body of literature.

Research question 2 also asks whether low attachment security is significantly predicted by breastfeeding duration, net of shared familial and genetic factors. The link between attachment security and breastfeeding is often taken for granted instead of empirically tested, as

220 it is often assumed that mother-infant dyads who initiate and continue breastfeeding will exhibit relatively strong levels of attachment. The results of the present analysis first revealed that 52% of the variance in low attachment security during toddlerhood is attributable to the shared environment, none of the variance is attributable to genetic factors, and the remaining variance

(48%) can be attributed to the nonshared environment and error. Upon further analysis, the results also indicated that a shorter duration of breastfeeding increases the risk of a low level of attachment security in offspring during toddlerhood. In short, there appears to be a relative deficit in the socioemotional development of toddlers who fail to initiate breastfeeding or who engage in breastfeeding for shorter durations. Again, these findings are robust to the influence of shared environmental and genetic factors, as twin differences in breastfeeding duration predicted their relative level of independently-rated attachment security to the same caregiver. A threshold effect of breastfeeding on low attachment security, however, was not detected. Put differently, monthly changes in breastfeeding duration appear to impact the level of attachment security during toddlerhood, regardless of when the child is ultimately weaned. There was no evidence that the benefits of breastfeeding for toddler attachment security are restricted to offspring who attain “long-term” status (i.e., 6 months of breastfeeding or more). Still, attaining long-term status does appear to have the added benefit of directly reducing the degree of externalizing behavior during kindergarten.

The third research question explores whether attachment security represents a mediating mechanism that links breastfeeding practices to externalizing behavior. The findings, again, are mixed. The results suggest that a lower level of attachment security during toddlerhood increases the likelihood of externalizing behavior during kindergarten, which corroborates prior work in this area of research (Groh et al., 2012; Kochanska & Kim, 2013; O’Connor et al.,

221 2012). In conjunction with the findings linking a short duration of breastfeeding to attachment security, these results, and the results of a subsequent path analysis, suggest that month-to-month changes in breastfeeding duration have indirect effects on externalizing behavior, despite having no direct effects on externalizing behavior. Nevertheless, attachment security failed to mediate the previously detected relationship between breastfeeding for less than 6 months and externalizing behavior, implying that the association between being breastfed for less than 6 months and an increased level of externalizing behavior is not explained by diminished attachment security during toddlerhood. These results highlight the possibility that role of breastfeeding in behavioral outcomes might be underestimated to the extent that breastfeeding effects operate through other developmental processes, such as attachment security.

Research question 3 also inquires about the potential confounding role of low birth weight in the relationship between breastfeeding practices and externalizing behavior. Infants who are born low birth weight tend to be more susceptible to poor nutrition outcomes, including the failure to engage in long-term breastfeeding (Quigley et al., 2012; Smith et al., 2003).

Moreover, low birth weight children are also at greater risk of conduct problems during childhood (Jackson & Beaver, 2015). Thus, it is possible that twins who significantly differ in their birth weight might be both differentially prone to breastfeed for a shorter duration and differentially prone to engage in externalizing behavior, which may render any link between a short duration of breastfeeding and externalizing behavior spurious. The results of DF analysis revealed that the positive effect of breastfeeding less than 6 months on externalizing behavior persists even after accounting for the birth weight of each twin. Thus, the analyses yielded no evidence that breastfeeding is only serving as a proxy for birth weight in this study.

222 The fourth research question expands the current literature that links breastfeeding to childhood conduct problems by exploring whether the influence of breastfeeding duration on externalizing behavior is moderated by indicators of genetic, perinatal, and socioemotional risk.

To be precise, this portion of the dissertation seeks to determine whether biosocial interactions involving nutritional, genetic, and/or developmental components are having a significant impact on the likelihood of externalizing behavior problems during kindergarten. The results, again, are not entirely consistent. In short, evidence of biosocial interactions emerge in some instances, whereas in other instances, no such evidence emerges. Three varieties of biosocial interactions were examined: 1) interactions between the breastfeeding variables and genetic risk 2) interactions between the breastfeeding variables and low attachment security and 3) interaction between breastfeeding variables and low birth weight.

First, the results were uniformly supportive of an interactive relationship between breastfeeding duration and genetic risk in the prediction of externalizing behavior. The findings consistently indicated that the externalizing behavior of children at high genetic risk of such behavior was much more likely to be influenced, for better or for worse, by breastfeeding practices, regardless of how such practices were measured. Thus, children with high levels of genetic risk who engaged in shorter durations of breastfeeding (whether exclusive or nonexclusive) exhibited some of the worst externalizing behavior among the children in the sample (with predicted externalizing scores at the 92nd – 94th percentile, net of covariates).

Nevertheless, children at high genetic risk of externalizing behavior appeared to exhibit noticeably low levels of externalizing behavior when they engaged in breastfeeding for very long durations. Breastfeeding duration, however, was significantly less consequential for the behavior of children with lower levels of genetic risk.

223 Despite the evidence of gene-environment interplay in the present study, no evidence for an interactive relationship between breastfeeding and low attachment security emerged. In short, low attachment security failed to moderate the relationship between breastfeeding practices and externalizing behavior, suggesting that breastfeeding effects on conduct problems are not significantly different for children with different levels of attachment security. Thus, while the duration of breastfeeding (measured in months) appears to influence the level of attachment security in offspring, the effects of breastfeeding duration on externalizing behavior do not appear to be contingent on the level of attachment security during toddlerhood. Some evidence did emerge, however, for the moderating role of low birth weight. More specifically, the risk of externalizing behavior incurred as a result of failing to breastfeed for at least 6 months was exacerbated for low birth weight children. Normal birth weight children who failed to reach the

6 month mark, however, scored relatively low on kindergarten externalizing behavior. These interactive effects, however, did not persist in the case of continuous or exclusive breastfeeding measures.

Research questions 5 through 8 move beyond nutrition during infancy to early childhood nutrition. The fifth research question specifically inquires about the ways in which children with especially poor eating practices are different from children with healthier eating practices. The results of several zero-order, bivariate analyses revealed a few general patterns. First, an overall unhealthy eating pattern was associated with higher levels of externalizing behavior at the bivariate level, particularly in the case of low fruit and vegetable consumption, and high sweets consumption. Second, children with poor overall diet were more likely to be nonwhite and to be corporally punished on a regular basis. Finally, children who scored high on the low diet quality measure came from a) households with lower income and maternal education levels and b)

224 families where regular meal time, food rules and eating routines were not as heavily emphasized.

Interestingly, the results revealed that these general patterns varied depending on the component of diet examined. For example, although overall dietary pattern was no associated with attachment security, children who scored the lowest on fruit and vegetable consumption tended to score lower on attachment security during toddlerhood. Alternatively, groups differing in the severity of their sweets consumption did not appear to significant differ in terms of household income and maternal education, despite the tendency for children with generally poor eating patterns to come from homes with lower income and education levels. Ultimately, while the exact risk factors associated with low diet quality are not entirely consistent across the components of the low diet quality composite measure, the most consistently related child and household covariates that are significantly associated with poor dietary practices are 1) low household income (sig. in 5 out of 7 models), 2) low maternal education (sig. in 5 out of 7 models), 3) nonwhite (sig. in 4 out of 7 models), 4) no family food rules (sig. in 4 out of 7 models), 5) corporal punishment (sig. in 3 out of 7 models), and 6) low attachment security (sig. in 3 out of 7 models). Clearly, demographic factors appear to be most closely related to diet quality, with some family processes also being somewhat related to diet quality.

The sixth research question extends the findings of the bivariate analyses to within- family, genetically informative tests of the relationship between preschool diet quality and externalizing behavior, partly because the results of the bivariate analyses revealed different degrees of demographic and familial risk across a low diet quality threshold (the 90th percentile).

Specifically, question 6 asks whether a low quality diet during preschool significantly increases the risk of externalizing behavior during kindergarten, net of familial and genetic influences.

Again, the results indicated that 83% of the variance in externalizing behavior is due to genes,

225 none of the variance in externalizing behavior is due to the shared environment, and the remaining variance (17%) in externalizing behavior is due to the nonshared environment and error. The results also yielded a general pattern in which several dietary components during preschool influence the development of externalizing behavior during kindergarten. To illustrate, the overall measure of low diet quality, in addition to low fruit consumption, high fast food consumption, high sweets consumption, and high salty snack consumption, are all associated with significantly higher levels of kindergarten externalizing behavior, independent of the influence of shared environmental and genetic factors. Thus, even same-sex twins from the same household who differ in their eating habits during preschool seem to differ significantly in their level of externalizing behavior by kindergarten.

The seventh research question inquires about whether any association between preschool diet and subsequent externalizing behavior is explained away by a) low attachment security and/or b) pre-existing externalizing behavior problems. The results revealed that, while the shared environment explained a significant portion of the variance in dietary factors (ranging from 54% to 89%), genetic factors did not. Furthermore, the results of several analyses jointly revealed that low attachment security a) had no discernible effect on any of the dietary components and b) failed to confound the previously detected relationship between preschool diet and externalizing behavior. These results are somewhat unexpected in light of a handful of recent studies supporting a significant association between attachment security and subsequent diet (Bost et al., 2014; Bozorgi, et al., 2014; for an excellent summary, see Lu et al., 2013).

However, such empirical tests were not within-family, genetically informative tests, which might explain the divergent results. These findings suggest that the effect of attachment security on diet quality detected in prior studies may be overestimated and/or spurious due to omitted

226 variable bias and/or poor measurement. Furthermore, the results suggest that the effects of preschool diet on externalizing behavior are not merely a function of preexisting socioemotional process at an earlier stage of the life course.

Although low attachment security did not confound the relationship between preschool dietary factors and externalizing behavior, pre-existing levels of externalizing behavior, in some cases, altered this relationship. As would be expected, externalizing behavioral patterns during preschool were predictive of externalizing behavioral patterns during kindergarten, providing evidence of significant stability in behavioral profiles of children over the course of 1-2 years.

Nevertheless, twins with poor eating behaviors during preschool, relative to their cotwins, still exhibited significantly higher levels of externalizing behavior during kindergarten, even after taking preexisting levels of externalizing behavior into account. To be precise, low diet quality

(index), low vegetable consumption, low fruit consumption, and high sweets consumption all increased the relative level of externalizing behavior during kindergarten, net of preschool externalizing behavior. Thus, while the components of the diet that matter most in the development of externalizing behavior change slightly depending on whether stability in externalizing behavior is modeled, the substantive results of these model point to the same general conclusion: preschool dietary factors, on the whole, significantly influence the risk of externalizing behavior during kindergarten.

The eighth research question builds upon prior literature by asking whether nutritional factors during preschool interact with genetic and socioemotional In short, the findings yield support for the EEA, since factors to predict externalizing behavior during kindergarten. The results generally indicate that neither genetic risk nor low attachment security moderate the relationship between lot diet quality and externalizing behavior. To be more precise, low

227 attachment security failed to significantly moderate the relationship between any of dietary components and externalizing behavior, suggesting that the influence of preschool diet on subsequent behavioral problems is likely not contingent on the degree of mother-child attachment security. Thus, regardless of whether the child experiences relatively low or high levels of mother-child attachment, an improvement in preschool dietary factors would still be expected to minimize externalizing behavior problems during kindergarten. Put differently, the risks of externalizing behavioral problems for young children with a poor quality diet are not limited to those children with a specific attachment history, but instead are relatively similar across levels of attachment security.

Similarly, genetic risk, in most instances, does not appear to moderate the relationship between preschool dietary factors and externalizing behavior during kindergarten. However, the results yielded one important exception: the influence of fast food consumption on externalizing behavior was limited to children who possessed a relatively high level of genetic risk. To be precise, children who both a) consumed fast food at least once a day and b) possessed the highest level of genetic risk were predicted to score at the 94th percentile of externalizing behavior during kindergarten. Conversely, when children with high genetic risk avoided fast food altogether, their externalizing behavior was predicted to be 13 percentile points lower.

Importantly, for children who possessed low genetic risk, fast food consumption had little to no bearing on their externalizing behavior score. The results suggest that, at least in the case of fast food consumption, a poor diet might place children with a genetic propensity towards externalizing behavior even more at-risk than they would be if they consumed a healthier diet.

The ninth research question asks whether infant and early childhood nutrition might be associated. More specifically, does a short duration of breastfeeding increase the likelihood of a

228 low quality diet during preschool? This question sets the stage for additional explorations of the interplay of nutritional factors in the prediction of externalizing behavior during kindergarten.

The results reveal that nutrition during infancy and early childhood are closely related, as children who exhibit the worst eating behaviors during preschool, regardless of the dietary component examined, are more likely to have breastfed for a shorter duration (or to have never initiated breastfeeding) than children who exhibit better eating habits. Importantly, the same cannot be said for exclusive breastfeeding: the worst eaters, across most dietary components, were not significantly less likely to have engaged in long-term exclusive breastfeeding than children with healthier diets, even though the worst eaters are clearly more likely to be weaned relatively early.

Of course, the clear follow up question is whether the link between nutrition across life stages is actually causal. Does breastfeeding protect against a poor quality diet later in life, or are both simply manifestations of certain shared environmental or genetic factors? The tenth research question asks specifically whether a low quality diet links a short duration of breastfeeding to externalizing behavior. The results first suggest that, independent of shared environmental and genetic factors, a shorter duration of breastfeeding increases the risk of a low quality diet and, more specifically, high salty snack consumption. Importantly, these effects did not persist when breastfeeding was modeled at the 6-month threshold and/or was modeled as exclusive breastfeeding. Thus, only month to month changes, irrespective of any threshold, influence dietary factors during preschool. In terms of mediation, the results only provide evidence of purely indirect effects of a shorter duration of breastfeeding on externalizing behavior through low diet quality. Thus, monthly reductions in breastfeeding duration result in a significant decline in diet quality during preschool, which in turn increases externalizing

229 behavior during kindergarten. Despite this finding, the robust, direct link between breastfeeding for less than 6 months and externalizing behavior that was detected previously was not explained away by preschool diet quality, regardless of which dietary component was examined. The results collectively reveal that dose-response effects of breastfeeding on externalizing behavior

(i.e., month-to-month) are in no way direct, but are purely indirect, as a reduction in the number of months breastfed significantly decreases attachment security and worsens diet quality, both of which increase the risk of externalizing behavior during kindergarten. Nevertheless, when examining threshold effects (i.e., breastfed less than 6 months), the influence of breastfeeding on externalizing behavior is direct and significant, with none of the examined mechanisms explaining this relationship.

The final and eleventh research question explores whether breastfeeding practices and early childhood diet interact to predict externalizing behavior during kindergarten. The results are somewhat mixed, but generally suggest little evidence of any consistent pattern of interactions between dietary components and breastfeeding measures (only 3 out of 16 interactions are significant). Nevertheless, there are a few important exceptions to this general rule. First, the results revealed that the influence of breastfeeding less than 6 months on externalizing behavior was significantly stronger for children who ate fewer vegetables during the preschool years. More precisely, eating no vegetables during the week prior to the preschool-year interview was especially influential in increasing the externalizing behavior of children who failed to breastfeed for at least six months. Second, the effect of breastfeeding less than 6 months on externalizing behavior during kindergarten was also significantly stronger for children who ate a high number of salty snacks around age 4. This interaction is similar to, and complements, the breastfeeding x low vegetable consumption interaction. Although greater

230 consumption of salty snacks did not increase the externalizing behavior of children who engaged in breastfeeding for 6 months or more, it did increase the externalizing behavior of children who failed to reach or surpass the 6-month threshold. Thus, at least in the case of low vegetable consumption and high salty snack consumption, a poor diet during early childhood may only correspond to higher levels of externalizing behavior among children who were not exposed to long-term breastfeeding during infancy.

Finally, the findings also revealed that the influence of sweets consumption on externalizing behavior was contingent on exposure to long-term, exclusive breastfeeding during infancy. To be precise, the amount of sweets consumed around age 4 emerged as more relevant to subsequent externalizing behavior of children who participated in long-term, exclusive breastfeeding, so much so that consuming no sweets v. consuming them daily increased the predicted externalizing behavior scores among exclusively breastfed subjects from the 12th percentile to the 74th percentile. Thus, sweets consumption is the only dietary factor that is especially capable of worsening the behavior of children exposed to long-term, exclusive breastfeeding. Nevertheless, the mixed pattern of findings regarding the interplay of nutritional factors across infancy and early childhood should still be interpreted with caution, as the general pattern of the results pertaining to research question 11 is one of null effects.

Ultimately, there are nine key conclusions that can be drawn from the collective body of findings presented in this summary. First, exposure to various nutritional factors during infancy and early childhood is not random, but instead is closely related to other indicators of environmental and demographic risk, from race to household income to maternal education.

Second, genetic factors explain a significant portion of the variance in externalizing behavior, despite failing to explain a significant portion of either low attachment security or low diet

231 quality. Ultimately, the relative influence of genetics, the shared environment, and the nonshared environment is distinct depending on the specific developmental process being examined. Third, the effects of breastfeeding on externalizing behavior may be direct, indirect, or null, depending on the way breastfeeding is measured. When the effects are indirect, both low diet quality and low attachment security appear to play an important role in linking a short duration of breastfeeding to externalizing behavior. Generally speaking, the exclusivity of breastfeeding appears to be mostly irrelevant to the development of externalizing behavior, while the exact timing of weaning seems to be directly relevant to the development of externalizing behavior. Fourth, the effect of breastfeeding, whether exclusive or nonexclusive, is highly contingent on the degree of genetic risk possessed by the child. Undoubtedly, the most robust biosocial process detected in the present study is the interaction between genetic risk and all three of the breastfeeding measures. Fifth, longer durations of breastfeeding appear to be more beneficial in curbing the externalizing behavior of low birth weight children.

Sixth, a number of nutritional factors during early childhood impact the degree of externalizing behavior exhibited during kindergarten, regardless of whether preexisting externalizing behavior or low attachment security is taken into account. Seventh, biosocial interactions between early childhood diet and other genetic and developmental processes were virtually nonexistent, despite the detection of an exacerbating effect of fast food on the relationship between genetic risk and externalizing behavior. Eighth, the duration of breastfeeding and subsequent diet during early childhood appear to be related, especially in the case of overall diet quality and salty snack consumption. Nevertheless, the effect of breastfeeding less than six months on externalizing behavior is not explained by low diet quality.

Ninth, dietary factors across the two life stages generally do not interact to predict externalizing

232 behavior, despite some evidence which suggests that breastfeeding for 6 months or more might protect against the deleterious effects of certain poor dietary practices (i.e., low vegetable consumption, high salty snack consumption) on childhood conduct problems.

7.2 Policy Implications

The results of the current study intimate a number of possible policy implications.

Broadly speaking, the results suggest that interventions aimed at preventing and/or minimizing antisocial behavior will likely prove more successful if they a) are implemented during earlier stages of the life course and b) take nutritional factors during infancy and early childhood into account. This is not to say that nutritional interventions during later stages of the life course are ineffective or fruitless (see Gesch et al., 2002; Zaalberg et al., 2010), or that nutritional factors are the only risk factors that should be targeted during childhood. However, comprehensive prevention efforts would likely benefit from incorporating nutritional elements at the earliest stages of life (Olds, 2006; see also Raine, Mellingen, Liu, Venables, & Mednick, 2003). Of course, prevention efforts will not always be available and/or effective, and so some children will still end up on an antisocial path regardless, making treatment programs for adolescents and adults indispensable. Notwithstanding, the benefits of intervening before a life-course trajectory of offending is in full swing cannot be understated. The current study bolsters the argument that that, when it comes to biosocial risk factors, such as nutrition, “early intervention is of paramount importance” (Rocque et al., 2012, p. 311).

The results also inform a number of more specific policy recommendations pertaining to both breastfeeding and early childhood nutrition. First and foremost, efforts to educate expectant mothers and fathers on the process of breastfeeding, as well as the potential benefits of

233 breastfeeding for offspring health and behavioral trajectories, are absolutely crucial. Such education is of particular relevance for low-income and minority populations, as these groups tend to either breastfeed for very short durations or avoid breastfeeding altogether. Programs such as these are already in place in a number of locations across the country. For example,

Nurse Family Partnership (NFP), a program which provides medical support (i.e., nurse home visits) and education for young and/or disadvantaged first-time moms from pregnancy through the toddler years, is currently operating in various counties across 43 states in the United States.

Programs like NFP do much in the way of educating mothers on how to facilitate a healthy pregnancy and how to best care for their infant following delivery. Research has illustrated that, on average, participation in nurse visitation programs like NFP results in a significantly longer duration of breastfeeding (see Kemp et al., 2011, for an example).

Despite the robust benefits of NFP for both the health and well-being of the mother and child (Olds, 2006), some states have taken steps to diminish the scale and scope of the program, leaving some of the most disadvantaged counties without access to the services it provides. For instance, in 2011, Rick Scott, the governor of Florida, cut nearly $2 million in health services for at-risk women and children, $500,000 of which had been previously appropriated to run NFP.

The cuts are estimated to leave hundreds of women and children without the health services they need. While it is too soon to determine the exact outcome of this specific cut-back in NFP services, the results of three large-scale, randomized control trials of the program suggest that high-risk children who are denied the program will likely have more arrests and convictions, greater emergent substance use, and an earlier-onset of sexual activity than if they had participated in the program (see Olds, 2006). The results of the current study provide some degree of support for NFP by illustrating that breastfeeding can reduce the risk of antisocial

234 behavior in children, at least in the short term. However, the results of the current study also suggest that breastfeeding education may need to get more specific, even tailor-made, to have its most potent effect on child outcomes. To be precise, the findings suggests that, while breastfeeding may have some direct and indirect effects on children’s behavior, the size of this effect is quite variable depending on the presence of other risk factors (e.g., genetic risk, perinatal risk, and poor early childhood diet). In short, even though the influence of breastfeeding on offspring antisocial behavior may not be particularly noteworthy across all mother-child dyads, it can be highly impactful for a select subset of mother-child dyads with other preexisting and/or concurrent risk factors. As a result, more nuanced knowledge about the exact benefits of breastfeeding should be made available to the public as it emerges so that parents are empowered to make the choice they feel is best for the health and well-being of their child.

Ultimately, the results of the current study imply that the risk incurred by a short duration of breastfeeding should be determined in the context of the risk incurred by other developmental factors. The question can no longer be as simple as “Does breastfeeding matter?” Instead, the question must evolve into “To what extent does breastfeeding matter for different segments on the population, stratified by known risk factors?” Answering these questions will likely maximize the benefit of any program that seeks to promote and facilitate long-term breastfeeding among mothers who currently forego extended breastfeeding durations.

Because the current study a) found some support for the role of long-term breastfeeding in protecting against offspring antisocial behavior and b) failed to detect substantial risk of a longer duration of breastfeeding to the behavioral profile of offspring, it is reasonable to implement policies that, at a minimum, do not interfere with the efforts of mothers to breastfeed

235 their infants. For example, employers should implement specific policies that permit breastfeeding employees to express (i.e., pump) milk during work breaks and meal times.

Furthermore, a safe, sanitary, and private space should be provided at work for expressing and storing breast milk. Employers, depending on the circumstances and resources available, might also provide or rent an electric breast pump to the breastfeeding employee during the duration of breastfeeding to enable the employee to continue breastfeeding for as long as she wishes. While these policies may seem self-evident and simple, they ensure that women who desire to breastfeed their infant(s) are not penalized simply because they desire and/or are required to work. Generally speaking, an environment of support in the workplace, where women are not stigmatized or ridiculed for their decision and efforts to breastfeed, should be encouraged.

The results of this study also have clear implications for hospital policy regarding breastfeeding that can be (and frequently are) adopted in an effort to encourage breastfeeding among mothers. For example, hospitals should consider making every effort to allow the mother and the infant to experience skin-on-skin contact immediately following the birth and to initiate breastfeeding as soon as possible after birth (e.g., within an hour). During her stay at a hospital or birthing center, mothers should be permitted to stay with their infant, except in the case of medical emergencies, in order to enable her to establish a pattern of breastfeeding. Lactation consultants/nurses should be on hand following the delivery of the baby to assist the infant in latching on to the breast and to aid in the general development of proper breastfeeding practices.

It is important to note that in no way do these policies remove the volition of the mother to refuse to breastfeed. Instead, the option of breastfeeding is made available to the patient, along with the information pertaining to the benefits of breastfeeding, so that the mother can make the choice for herself. Similar scenarios emerge in medical settings regarding other medical procedures,

236 from surgeries to flu shots: staff provide the opportunity and the knowledge, but patients ultimately make the decision. The current study yielded no findings that would suggest that offering such education and services is irresponsible and/or undesirable. On the contrary, doing so can only help mothers in their quest to breastfeed, which at best proves substantially beneficial for the child, and at worst has little to no effect on the development of the child.

Finally, some of the results suggest that efforts to encourage breastfeeding should be prioritized for infants who experience certain perinatal risk factors around the time of birth (e.g., low birth weight). Since breastfeeding appears to be more consequential for the behavior of low birth weight children, policies should be implemented to ensure that such children are provided with breast milk if the mother so desires. There can be added challenges to this process if the infant is experiencing significant health problems and/or is receiving intensive care for an extended period of time. Nevertheless, feeding the infant expressed milk during this time, as opposed to formula, may yield some additional benefits in terms of subsequent behavior, and thus should be encouraged when possible. On some occasions, mothers of low birth weight infants may not be capable of producing sufficient milk, and may need to rely on donor milk to feed their infant while he/she is in the hospital receiving care. Such options should be provided by the hospital, and additional resources needed to initiate and/or reestablish breastfeeding once the health of the infant has stabilized should also be provided. Still, it is important to acknowledge that the full impact of breastfeeding on toddler attachment (which itself influences externalizing behavior) may not emerge if feeding at the breast is not eventually implemented.

Similarly, the results suggest that children with a high level of genetic risk of externalizing behavior problems are most sensitive to the duration of breastfeeding. When it comes to challenging behavioral outcomes, these children appear to benefit the most from long-

237 term breastfeeding, but also suffer the most when it is not provided. Policy researchers often refer to this differential response to risk as the responsivity principle of effective intervention

(Bonta & Andrews, 2007). In short, characteristics of individuals can alter their capacity to adequately benefit from an intervention. In the case of breastfeeding, it appears that both genetic and perinatal information might provide clues about whether, and to what extent, breastfeeding should be strongly encouraged among mother-infant pairs. Of course, research examining genetic sensitivity to breastfeeding is only in its infancy, and the current study represents one of the first to explore the moderating role of genetic risk in the relationship between breastfeeding practices and externalizing behavior. Even so, because the current study measures genetic risk latently, the primary obstacle to implementing a policy that takes genetic risk into account is that the specific plasticity markers (i.e., risk alleles) that influence behavioral sensitivity to breastfeeding remain unidentified. Upon identifying these markers, they could be used to determine the likelihood that the child’s behavior will be substantially influenced by the decision to breastfeed long term. Although currently implementing such a policy would certainly be premature, the findings of the present study represent an important first step on the path to genetically tailored prevention efforts. Similar to the field of medicine, genetically informed prevention efforts would likely prove more effective and efficient than other prevention efforts, in part because they would be rooted in a proven principle of effective intervention: the responsivity principle (Bonta & Andrews, 2007; Polaschek, 2011).

In terms of early childhood diet, the current study is consistent with the efforts of nutritionists worldwide to encourage healthy eating among the global population. Moreover, the findings add to the numerous studies of young children, suggesting that what children eat is associated with how they behave. The findings, however, give even more reason to believe that

238 this relationship is actually causal, as results were robust to both shared environmental and genetic factors. Furthermore, the findings indicate that changes to the diet during early childhood can have relatively long-lasting effects on behavior. Consequently, every effort should be made, at both home and school, for children to be provided with a diet characterized by a greater number of fruits and vegetables, and few sweets, salty snacks, and fast food items.

Clearly, educating parents, teachers, school administrators and even young children about the relevance of diet to the behavioral patterns of children is essential. For example, required nutrition classes could be featured on the curriculum of schools in order to encourage healthy eating among students.

Of course, additional steps can be taken to ensure better eating patterns. One step is to provide healthy, balanced meals to disadvantaged children through a free or discounted lunch program. This is currently in place across countless locales in the United States, and should be continued, but with a greater emphasis on the nutritional value of the food served. It is of particular importance that economically disadvantaged children be provided with a nutritional lunch at school, as they are less likely to be receiving nutritious food on a regular basis at home.

Still, as the results revealed, nutritional factors during early childhood appear to impact the behavior of children across social class, since twins in the same household with differing eating habits tended to diverge in their behavioral patterns. Therefore, steps should be taken during elementary school to provide children and parents with simple, easy-to-understand labels that summarize the nutritional value of the foods offered at school. Although this appears daunting, streamlined labels of the nutritional quality of food, created by experts in medicine/nutrition and rooted in rigorous empirical research, are already available and could be utilized by schools nationwide (see Katz, Njike, Rhee, Reingold, & Ayoob, 2010). By providing healthy meals for

239 the economically disadvantaged students, supplying more healthy options for children who purchase lunch at school, and incorporating adequate education on the impact of nutrition on health, cognition, and behavior into the school curriculum, families and children will likely improve their eating behaviors or, at a minimum, will be better equipped to make more informed choices about what they are eating.

Finally, some of the findings of the current study indicated that children may be differentially susceptible to the effects of poor diet on behavior depending on their level of genetic risk and their prior exposure to breastfeeding. Although the specific genetic plasticity markers that might make the behavior of children more responsive to diet remain unknown, it is quite easy to establish the breastfeeding history of the child and, as a result, implement a nutritional intervention that would likely be most helpful to them. Additional findings suggest that certain dietary interventions may be more effective at reducing behavioral problems if more of the effort is direct at 1) children with genetic risk for behavioral problems and/or b) children with little to no breastfeeding history. Still, these results were not consistent across dietary factors, and so should be interpreted with caution.

7.3 Limitations of the Current Study

Although the current study empirically examines of number of previously unexplored questions regarding the relationship between nutritional factors and childhood antisocial behavior, it is not without its shortcomings. Specifically, there are five key limitations to this dissertation.

First, the analytical sample of the current study only consisted of MZ and DZ twins. Of course, restricting the sample to twins was necessary in order to empirically examine many of the

240 research questions proposed at the beginning of this document. Nevertheless, it is possible that the associations detected may not hold in the singleton population, or that some of the key variables of interest (e.g., breastfeeding) may be uniquely common or uncommon in the twin population. Ultimately, the only key variables of interest that varied between twins and singletons at any substantive rate were low birth weight and long-term breastfeeding, with singletons being significantly less likely to be born low birth weight (57% v. 20%; t < .00) and significantly more likely to breastfeed long term (19% v. 29%, t < .00). Strictly speaking, therefore, the results of the current analysis cannot be generalized to the singleton population.

Despite this weakness, there are numerous strengths in the current research design that other designs lack, including the ability to account for genetic and shared environmental factors when testing hypotheses and to test for GxE. Furthermore, a number of researchers have found that studies employing large, nationally representative samples of twins tend to be more generalizable to the singleton population than frequently assumed (for an example, see Barnes & Boutwell,

2013).

Second, additional breastfeeding details would have been preferred, but such details were not provided in the ECLS-B data. For example, it would have been particularly useful to know whether, and to what extent, mothers fed their infant expressed breast milk or milk directly from the breast. Although this may seem like a superfluous distinction, data on these details would have enabled me to tease apart the effects of breast milk as a nutritional factor and the effects of breastfeeding as a social experience. It would permit a more nuanced examination of exactly how breastfeeding impacts attachment, and whether expressed breast milk has similar or different benefits for behavior relative to feeding directly from the breast. Although the breastfeeding measures available in the ECLS-B are not incredibly detailed, they do permit the

241 comparison of exclusive versus nonexclusive breastfeeding, due to the inclusion of variables such as those concerning the onset of formula feeding and solid-food feeding.

Third, the measures of early childhood nutrition were somewhat limited in their scope and specificity. That is, when examining the influence of early childhood nutrition on externalizing behavior, it would have been worthwhile to also test hypotheses using measures that tap specific nutrients (e.g., omega-3 fatty acids, iron), but such measures were not available in the data. Some of the recent nutrition literature utilizes very precise measures of nutrient intake, such as omega-3 blood levels (Gow et al., 2013) and iron deficiency (Golub & Hogrefe,

2014). Nevertheless, while these would be preferable at the molecular and/or neurological level, more generalized measures of poor diet are useful in highlighting the specific dietary changes that can be made on a daily basis to minimize behavioral problems (see Park et al., 2012; Oh et al., 2013).

Fourth, while the ECLS-B contains a large sample of twins, and is fairly comprehensive in scope, subjects were only followed until they were approximately 6 years of age. From a criminological perspective, it would have been informative to follow these children into adolescence and adulthood in order to examine their delinquency trajectories. The data provide no means of explicitly testing for the development of criminal behavior, despite the wealth of data relating to the development of antisocial behavior during childhood. Despite this important limitation, the ECLS-B data are incredibly thorough in their coverage of multiple aspects of child development, as well as numerous prenatal, perinatal and early childhood risk factors. To my knowledge, no nationally representative dataset examining delinquent and criminal behavior provides such rich detail concerning the first few years of life as the ECLS-B, which is why it was chosen for the present study. Furthermore, research has revealed that identifying a class of

242 individuals who have severe behavioral and self-regulation problems during childhood is one of the best ways to predict life-course-persistent offending (Vaughn, DeLisi, Beaver, & Wright,

2009). Importantly, breastfeeding risk and genetic risk, in conjunction, placed children in the top

6-8% of the externalizing distribution, suggesting that they are indeed at risk of life-course- persistent offending. In this way, a dataset that extends beyond childhood may not be as germane to the study of criminological processes as is often presumed. Still, as research similar to the current project becomes more commonplace in the field of criminology, more comprehensive studies that span from birth to adulthood will hopefully be conducted, enabling a more complete test of the research questions relevant to life-course criminology.

Finally, the ECLS-B data do not allow for tests of specific gene-environment interactions between the nutritional factors and measured polymorphisms (i.e., plasticity markers). While the latent measure of genetic risk that was employed in the current study is a valuable first step in examining gene-environment interactions, ultimately, interactions between nutritional factors, such as breastfeeding, and specific risk alleles (e.g., MAOA, DRD2, 5HTTLPR) should be identified, as doing so provides more of a basis for genetically informed interventions.

Importantly, some research had already examined the moderating effects of FADS2 in the relationship between breastfeeding and cognition. Still, breastfeeding research examining other genes as moderators, and/or behavioral outcomes, are completely absent from the literature

(except see Jackson & Beaver, in press). Studies exploring gene-diet interactions as predictors of behavioral problems during childhood, moreover, are essentially nonexistent. While the current study was unable to explicitly test risk alleles as moderators, the findings provide some initial evidence of the moderating role of genetic risk and suggest that significant interactions between specific genes and nutritional factors are likely occurring.

243 7.4 Future Directions for Theory and Research

The findings of this dissertation have ultimately highlighted the utility of research that explores the mediating and moderating role of genetic, socioemotional, and other developmental mechanisms in the link between nutrition and externalizing behavior. So much focus has been placed on straightforward, direct relationships between nutrition and behavior that it seems that most researchers have not considered whether the impact of nutrition on behavior may at times be obscured by assuming a unilateral response to nutritional intake across individuals within a population. Such an assumption is tantamount to presuming that a new pharmaceutical will be equally effective for all patients, or that enhanced physical activity will result in the same degree of weight loss for all individuals who participate in it. Ultimately, variation in the sensitivity to environments is absolutely germane to the issue of breastfeeding and early childhood nutrition, yet it is rarely acknowledge in this body of literature (particularly in the breastfeeding literature).

In the case of infant nutrition, most researchers do not fully consider how nutrition may indirectly influence behavior well into the future, and thus do not explore relevant mediating mechanisms across multiple development stages. Moving forward, researchers need to start explaining why, and under what conditions, nutritional factors might impact externalizing behavior, instead of prolonging the debate about whether they matter at all. For example, it is fairly clear that the direct effects of breastfeeding on behavior have been overstated due to the methodological limitations of most of the literature (see Colen & Ramey, 2014). Still, scholars have yet to fully explore the extent to which divergent results are the product of a heterogeneous response to breastfeeding. Moreover, scholars have not adequately addressed whether breastfeeding might lead to the development of various risk factors during toddlerhood and/or childhood, which in turn place children on an antisocial trajectory. It is not unreasonable to

244 expect that examining the influence of breastfeeding on behavior multiple years into the future might yield null results; it is unreasonable, however, to examine unilateral, direct effects, find null results, and then assume that infant nutrition is completely unrelated to behavioral problems in the long term. Scholars need to start theorizing about the processes and the conditions relevant to the nutrition-externalizing relationship, instead of continually trying to prove or disprove the existence of a universally applicable association.

Ultimately, this body of research would benefit from expanding the scope of possible mediating mechanisms linking a short duration of breastfeeding to externalizing behavior. The current study examined only two: low attachment security and low diet quality. Surprisingly, mediating mechanisms have generally been overlooked in the literature. While scholars make a number of assumptions about why breastfeeding would impact behavior, we have little research that provides an explanation of the frequently detected association. Future research also needs to expand upon the measure of genetic risk employed in the current study and test for pertinent gene-environment interplay between nutritional factors and measured genetic polymorphisms.

Doing so will build upon the findings of this study and may lead to more effective interventions involving nutritional elements.

Finally, criminologists need to consider the distinct possibility that the theoretical and empirical growth of the discipline may be stunted by needless adherence to artificial disciplinary boundaries. The developmental processes during infancy and early childhood that set the stage for physical and mental health, personality, cognition, and behavior are too numerous to count, and yet, as a discipline, we are often reluctant to incorporate life-course findings from myriad disciplines into mainstream criminological theory. On a related note, complex, biosocial interplay between biological/genetic and environmental factors is still seen as peripheral to

245 criminological theory, despite the frequency with which such interplay is detected and the clear relevance such interplay has for criminological theory (Belsky & Beaver, 2011; Liu & Wuerker,

2005; Raine, 2002).

Ultimately, the findings of the current study provide support for a biosocial approach to the study of antisocial behavior, as it was consistently revealed that children characterized by both a high-risk nutritional profile during infancy and a high-risk genetic profile were predicted to exhibit some of the most severe externalizing behavioral patterns during kindergarten. Such findings would have been completely overlooked had a standard social science approach been employed in this study instead of a modeling strategy that allows for the empirical testing of gene-environment interplay. Unfortunately, as long as criminologists continue to downplay the relevance of developmental and biosocial processes in the formation of antisocial behavior, they will continue to employ standard social science methodologies that are incapable of adequately examining such processes and, as a result, will be missing an integral component of the complex underpinnings of crime and delinquency.

246 APPENDIX A

ITEMS FOR NUTRITION MEASURES

Breastfeeding Items (Waves 1 and 2)

1. CD005: Did {you/{CHILD/TWIN}’s mother} ever breast-feed {CHILD/TWIN}?

2. CD015: {Are you/Is {CHILD/TWIN}’s mother} still breast-feeding {CHILD/TWIN}

now?

3. CD020: For how many months did {you/{CHILD/TWIN}’s mother} breast-feed

{him/her}?

Breastfeeding Exclusivity Items (Waves 1 and 2)

4. CD030: How old was {CHILD/TWIN} in months when you began feeding {him/her}

formula?

5. CD040: How old was {CHILD/TWIN} in months when you began feeding {him/her}

cow’s milk?

6. CD045: How old was {CHILD/TWIN} in months when solid food was first introduced?

Solid foods include cereal and baby food in jars, but not finger foods.

Low Diet Quality Items (Wave 3)

1. CH043: During the past 7 days, how many times did your child drink Soda pop (for

example, Coke, Pepsi, or Mountain Dew), sports drinks (for example, Gatorade), or fruit

drinks that are not 100% fruit juice (for example, Kool-Aid, Sunny Delight, Hi-C,

Fruitopia, or Fruitworks)?

2. CH044: During the past 7 days, how many times did your child eat fresh fruit, such as

apples, bananas, oranges, berries or other fruit such as applesauce, canned peaches,

canned fruit cocktail, frozen berries, or dried fruit? Do not count fruit juice.

247 3. CH045: During the past 7 days, how many times did your child eat vegetables other than

French fries and other fried potatoes? Include vegetables like those served as a stir fry,

soup, or stew, in your response.

4. CH046: During the past 7 days, how many times did your child eat a meal or snack from

a fast food restaurant with no wait service such as McDonald’s, Pizza Hut, Burger King,

Kentucky Fried Chicken, Taco Bell, Wendy’s and so on? Consider both eating out, carry

out, and delivery of meals in your response.

5. CH047: During the past 7 days, how many times did your child eat candy (including Fruit

Roll-Ups and similar items), ice cream, cookies, cakes, brownies, or other sweets?

6. CH048: During the past 7 days, how many times did your child eat potato chips, corn

chips such as Fritos or Doritos, Cheetos, pretzels, popcorn, crackers or other salty snack

foods?

248 APPENDIX B

ITEMS FOR THE LOW ATTACHMENT SECURITY MEASURE

Hot Spot 1: Is Comfortably Cuddly (Wave 2)

1. Hugs and cuddles against parent without being asked to do so.

2. Relaxes when in contact with parent.

3. Seeks and enjoys being hugged by parent.

4. When crying or upset, is easily comforted by contact with parent.

Hot Spot 2: Is Cooperative (Wave 2)

1. When parent asks child to do something, child understand what she wants—may or may

not obey.

2. Cooperates with parent and gives parent things if asked.

3. Responds to positive hints from parent.

4. Obeys when asked to bring or give parent something.

5. When parent says “come here”, child obeys.

Hot Spot 3: Enjoys Company (Wave 2)

1. If asked, lets friendly adult strangers hold or share toys.

2. A social child who enjoys the company of others.

3. Enjoys being hugged or held by friendly adult strangers.

4. Eager to join in with friendly adult strangers—does not wait to be asked.

5. Enjoys copying what friendly adult strangers do.

Hot Spot 4: Is Independent (Wave 2)

1. Often plays out of parent’s sight.

2. Is very independent.

249 3. Fearless, gets into everything.

4. Usually finds something else to do when finished with an activity—does not go to parent

for help.

5. Takes off and explores new things on own.

6. Hardly ever goes to parent for any help, not even for minor injuries.

Hot Spot 5: Seeks Attention (Wave 2)

1. Tries to stop parent from giving affection to other people, including family members.

2. When parent talks with anyone else, child wants parent’s attention.

3. Wants to be at the center of parent’s attention.

4. When child is bored, will go to parent looking for something to do.

5. Child frequently wants parent’s attention.

Hot Spot 6: Is Upset by Separation (Wave 2)

1. Is very clingy, stays closer to parent or returns more often than simply keeping track of

parent’s whereabouts.

2. Gets upset if parent leaves and shifts to another place.

3. Child does not try new things and always wants parent to help.

4. Cries or tries to stop parent from leaving or moving to another place.

Hot Spot 7: Avoid Others/Not Sociable (Wave 2)

1. Soon loses interest in friendly adult strangers.

2. Turns away from friendly adult strangers and goes own way.

3. If there is a choice, child prefers to play with toys rather than with friendly adults.

4. When a new visitor arrives, child first ignores or avoids him/her.

250 Hot Spot 8: Is Demanding/Angry (Wave 2)

1. When child cries, cries loud and long.

2. When child sees something really nice to play with, child will fuss and whine or try to

drag parent over to it.

3. When parent does not do what child wants right away, child fusses, gets angry or gives

up.

4. Cries as a way of getting parent to do what is wanted.

5. Cries often, regardless of how hard or how long.

Hot Spot 9: Is Moody/Unsure/“Unusual” (Wave 2)

1. Seems lost, remote and/or disconnected.

2. With parent, child suddenly changes moods—may go from being calm to upset, afraid or

angry, or from nice to mean, or gets upset and then goes bank. Goes all floppy (limp)

when held by parent.

3. Child sometimes freeze, perhaps in an unusual position, for a few seconds.

4. Will go towards parent to give parent toys, but does not touch nor look at parent.

5. Suddenly aggressive towards parent for no reason (for example, hits, slaps, pushes, or

bites parent).

6. Easily become angry at parent.

The exact item used as an overall measure of low attachment security in the current study is

X2TASSEC. This item is the traditional security factor score obtained using the same method that was originally employed by Waters and Deane (1985). This method is referred to as criterion sorting. Waters and Deane (1985) describe the process as follows:

251 Criterion sorting.- Judges can use a Q-set to operationalize important attachment constructs by sorting the items to describe a hypothetically most secure, dependent, or sociable subject.

Item-by-item comparisons between the placement of items by sorters defining one construct and sorters using the same items to define a different construct can be used to evaluate similarities and differences among related constructs (e.g., Deane & Waters, 1984; Waters,

Noyes, Vaughn, & Ricks, in press).

When constructs have been defined in this way, subjects can be scored on each construct by computing the correlation between the composite description of the "hypothetically most x subject" and the Q-sort description of a particular subject (i.e., the correlation between two arrays of scores within each subject). The correlation coefficient between the construct definition and the description of the subject is used as the subject's score on that construct.

The more similar the subject is to the hypothetically most extreme subject, the higher the subject's score on the construct.

This procedure has several important advantages. First, it enables us to place some distance between the observers who collect the primary data in a study and the constructs that will be scored from their data. Biases and halo effects are much less likely to intrude when observers use a Q-set to "describe a subject's behavior" without reference to any specific constructs than when observers are asked to assign ratings on the constructs themselves.

Second, it allows us to employ experts' definitions of a construct to score subjects without having to enlist the experts as observers.

The criterion-sort procedure also insures that the full range of relevant behavior is considered in assigning scores on a construct. Most constructs have implications for a wide range of behaviors. And in principle a high score should be reserved for subjects who have a broad

252 profile of construct-relevant behavior. In practice, raters who have a particular construct in mind respond strongly to positive evidence (e.g., to a few clear signs of insecurity in a particular interaction). As a result, moderate to high scores are often assigned to subjects whose behavior is unexceptional in much of the domain relevant to the construct. In contrast, a Q-sorter's task is to describe the subject's behavior with equal attention to every

Q-set item. After the subject has been described in detail, a high correlation between this description and a criterion sort implies exceptional behavior across a significant range of construct-relevant behavior. Isolated events are much less likely to result in high scores in

Q-sort data than they would be with conventional rating methods.

Finally, this procedure enables us to develop criterion sorts and assign scores on new variables long after data collection has been completed. This is a great advantage in longitudinal research. Interpreting unexpected results or alternative hypotheses is often facilitated by the ability to score subjects on a variable for which no specific measure was included in the study (Waters & Deane, 1985, pp. 55-56).

253 1 0.8 0.6 0.4 0.2 0 -0.2 Secure Attachment Prototype

Hotspot Score Score Hotspot -0.4 -0.6 -0.8 -1 -1.2 1 2 3 4 5 6 7 8 9

Hotspot Key: 1 = Comfortably Cuddly. 2 = Cooperative. 3 = Enjoys Company. 4 = Independent. 5 = Attention Seeking. 6= Upset by Separation. 7= Avoids Others. 8 = Demanding/Angry. 9 = Moody/Unsure/Unusual.

Figure B.1: Illustration of the Profile of the Secure Attachment Prototype Across Hotspots30

30 This figure was adapted from a figure prepared in 2005 by the developer of the TAS-45, Dr. John Kirkland of Massey University, Palmerston North, New Zealand.

254

1.2 1 0.8 0.6 0.4 0.2 0 Insecure Attachment Prototype -0.2 -0.4

Hotspot Score (Reverse-Coded) (Reverse-Coded) Score Hotspot -0.6 -0.8 -1 1 2 3 4 5 6 7 8 9

Figure B.2: Illustration of the Profile of the Insecure Attachment Prototype Across Hotspots

255 APPENDIX C

ITEMS FOR THE EXTERNALIZING BEHAVIOR MEASURE

Parent-Rated Items (Waves 4 and 5)

Next, I have some questions about {CHILD/TWIN}’s behavior. For each of the behaviors I read to you, I’d like you to tell me how often you see {CHILD/TWIN} behave in this way: never, rarely, sometimes, often, or very often. Whenever I ask you about how

{CHILD/TWIN} behaves with other children, consider children who are close in age to

{CHILD/TWIN} – no more than 2 years older or younger than {CHILD/TWIN}. Please base your answers on what you have seen of {CHILD/TWIN}’s behavior during the last 3 months.

How often in the last 3 months have the following things occurred? {CHILD/TWIN}:

1. Is physically aggressive, for example hits, kicks, or pushes.

2. Gets Angry (Exact wording copyrighted).

3. Is overly active – unable to sit still.

4. Has temper outbursts or tantrums.

5. Destroys things that belong to others.

6. Bothers and annoys other children.

7. Acts impulsively without thinking, for example runs across the street without looking.

Teacher-Rated Items (Waves 4 and 5)

For each of the behaviors, indicate how often you see the child behave in this way. For items that ask about how the child behaves with other children, consider other children in the class. [Response Options: Never(1), Rarely(2), Sometimes(3), Often(4), and Very Often(5)].

1. Is physically aggressive (for example, hits, kicks, or pushes).

2. Disrupts other children’s ongoing activities.

256 3. Is overly active – unable to sit still.

4. Has temper outbursts or tantrums.

5. Bothers and annoys other children.

6. Acts impulsively without thinking (for example, runs across the street without looking).

257 APPENDIX D

IRB APPROVAL FORMS

258

259

260 261 REFERENCES

Aarts, F., Geenen, R., Gerdes, V. E., van de Laar, A., Brandjes, D. P., & Hinnen, C. (2015). Attachment anxiety predicts poor adherence to dietary recommendations: An indirect effect on weight change 1 year after gastric bypass surgery. Obesity Surgery, 25(4), 666- 672.

Abraham, E. C., Godwin, J., Sherriff, A., & Armstrong, J. (2012). Infant feeding in relation to eating patterns in the second year of life and weight status in the fourth year. Public Health Nutrition, 15(09), 1705-1714.

Ainsworth, M. S. (1979). Infant–mother attachment. American Psychologist, 34(10), 932-937.

Alink, L. R., Mesman, J., Van Zeijl, J., Stolk, M. N., Juffer, F., Koot, H. M., ... & Van IJzendoorn, M. H. (2006). The early childhood aggression curve: Development of physical aggression in 10 to 50 month old children. Child Development, 77(4), 954-966.

Anderson, J. W., Johnstone, B. M.,‐ & Remley,‐ ‐ D. T. (1999). Breast-feeding and cognitive development: a meta-analysis. The American Journal of Clinical Nutrition, 70(4), 525- 535.

Anderson, S. E., Gooze, R. A., Lemeshow, S., & Whitaker, R. C. (2012). Quality of early maternal–child relationship and risk of adolescent obesity. Pediatrics, 129(1), 132-140.

Anderson, S. E., & Whitaker, R. C. (2011). Attachment security and obesity in US preschool- aged children. Archives of Pediatrics & Adolescent Medicine, 165(3), 235-242.

Andreassen, C., & Fletcher, P. (2007). Early Childhood Longitudinal Study, Birth Cohort (ECLS-B): Psychometric Report for the 2-Year Data Collection. Methodology Report. NCES 2007-084. National Center for Education Statistics.

Armstrong, J., & Reilly, J. J. (2002). Breastfeeding and lowering the risk of childhood obesity. The Lancet, 359(9322), 2003-2004.

Arseneault, L., Moffitt, T. E., Caspi, A., Taylor, A., Rijsdijk, F. V., Jaffee, S. R., ... & Measelle, J. R. (2003). Strong genetic effects on cross situational antisocial behaviour among 5 year old children according to mothers, teachers, examiner observers, and twins’ self reports. Journal of Child Psychology and Psychiatry‐ , 44(6), 832-848. ‐ ‐ ‐ Arseneault, L., Tremblay,‐ R. E., Boulerice, B., & Saucier, J. F. (2002). Obstetrical complications and violent delinquency: Testing two developmental pathways. Child Development, 73(2), 496-508.

262 Ashford, J., Van Lier, P. A., Timmermans, M., Cuijpers, P., & Koot, H. M. (2008). Prenatal smoking and internalizing and externalizing problems in children studied from childhood to late adolescence. Journal of the American Academy of Child & Adolescent Psychiatry, 47(7), 779-787.

Atkinson, S. A., Bryan, M. H., & Anderson, G. H. (1981). Human milk feeding in premature infants: protein, fat, and carbohydrate balances in the first two weeks of life. The Journal of Pediatrics, 99(4), 617-624.

Allen, J. P., Marsh, P., McFarland, C., McElhaney, K. B., Land, D. J., Jodl, K. M., & Peck, S. (2002). Attachment and autonomy as predictors of the development of social skills and delinquency during midadolescence. Journal of Consulting and Clinical Psychology, 70(1), 56.

Baker, L. A., Jacobson, K. C., Raine, A., Lozano, D. I., & Bezdjian, S. (2007). Genetic and environmental bases of childhood antisocial behavior: A multi-informant twin study. Journal of Abnormal Psychology, 116(2), 219.

Bakermans-Kranenburg, M. J., Van Ijzendoorn, M. H., & Juffer, F. (2003). Less is more: Meta- analyses of sensitivity and attachment interventions in early childhood. Psychological Bulletin, 129(2), 195-215.

Bakermans Kranenburg, M. J., & Van IJzendoorn, M. H. (2006). Gene environment interaction of the dopamine D4 receptor (DRD4) and observed maternal insensitivity predicting externalizing‐ behavior in preschoolers. Developmental Psychobiology‐ , 48(5), 406-409.

Ballard, O., & Morrow, A. L. (2013). Human milk composition: Nutrients and bioactive factors. Pediatric Clinics of North America, 60(1), 49-74.

Barnes, J. C. (2013). Analyzing the origins of life-course-persistent offending: A consideration of environmental and genetic influences. Criminal Justice and Behavior, 40(5), 519.

Barnes, J. C., Beaver, K. M., & Boutwell, B. B. (2011). Examining the genetic underpinnings to Moffitt's developmental taxonomy: A behavioral genetic analysis. Criminology, 49(4), 923.

Barnes, J. C., & Boutwell, B. B. (2013). A demonstration of the generalizability of twin-based research on antisocial behavior. Behavior Genetics, 43(2), 120-131.

Barnes, J. C., Boutwell, B. B., Beaver, K. M., & Gibson, C. L. (2013). Analyzing the origins of childhood externalizing behavioral problems. Developmental Psychology, 49(12), 2272.

Barnes, J. C., Wright, J. P., Boutwell, B. B., Schwartz, J. A., Connolly, E. J., Nedelec, J. L., & Beaver, K. M. (2014). Demonstrating the validity of twin research in criminology. Criminology, 52(4), 588-626.

263 Barone, L., & Lionetti, F. (2012). Attachment and emotional understanding: A study on late adopted pre-schoolers and their parents. Child: Care, Health and Development, 38(5), 690-696. ‐

Bates, J. E., Pettit, G. S., Dodge, K. A., & Ridge, B. (1998). Interaction of temperamental resistance to control and restrictive parenting in the development of externalizing behavior. Developmental Psychology, 34(5), 982.

Beaver, K. M. (2008). The interaction between genetic risk and childhood sexual abuse in the prediction of adolescent violent behavior. Sexual Abuse: A Journal of Research and Treatment.

Beaver, K. M., Barnes, J. C., May, J. S., & Schwartz, J. A. (2011). Psychopathic personality traits, genetic risk, and gene-environment correlations. Criminal Justice and Behavior, 38(9), 896-912.

Beaver, K. M., Boutwell, B. B., Barnes, J. C., Schwartz, J. A., & Connolly, E. J. (2014). A quantitative genetic analysis of the associations among language skills, peer interactions, and behavioral problems in childhood: Results from a sample of twins. Merrill-Palmer Quarterly, 60(2), 142-167.

Beaver, K. M., DeLisi, M., Vaughn, M. G., Wright, J. P., & Boutwell, B. B. (2008). The relationship between self control and language: Evidence of a shared etiological pathway. Criminology, 46(4), 939-970. ‐ Beaver, K. M., Schutt, J. E., Boutwell, B. B., Ratchford, M., Roberts, K., & Barnes, J. C. (2009). Genetic and environmental influences on levels of self-control and delinquent peer affiliation results from a longitudinal sample of adolescent twins. Criminal Justice and Behavior, 36(1), 41-60.

Beaver, K. M., & Wright, J. P. (2005). Evaluating the effects of birth complications on low self- control in a sample of twins. International Journal of Offender Therapy and Comparative Criminology, 49(4), 450-471.

Beaver, K. M., Wright, J. P., DeLisi, M., Walsh, A., Vaughn, M. G., Boisvert, D., & Vaske, J. (2007). A gene× gene interaction between DRD2 and DRD4 is associated with conduct disorder and antisocial behavior in males. Behavioral and Brain Functions, 3(30.10), 1186.

Beck, J. E., & Shaw, D. S. (2005). The influence of perinatal complications and environmental adversity on boys’ antisocial behavior. Journal of Child Psychology and Psychiatry, 46(1), 35-46.

Beitchman, J., Baldassarra, L., Mik, H., De Luca, V., King, N., Bender, D., ... & Kennedy, J. (2006). Serotonin transporter polymorphisms and persistent, pervasive childhood aggression. American Journal of Psychiatry, 163(6), 1103-1105.

264 Belfort, M. B., Rifas-Shiman, S. L., Kleinman, K. P., Guthrie, L. B., Bellinger, D. C., Taveras, E. M., ... & Oken, E. (2013). Infant feeding and childhood cognition at ages 3 and 7 years: Effects of breastfeeding duration and exclusivity. JAMA Pediatrics, 167(9), 836-844.

Bellisle, F. (2004). Effects of diet on behaviour and cognition in children. British Journal of Nutrition, 92(S2), S227-S232.

Belsky, J., & Beaver, K. M. (2011). Cumulative genetic plasticity, parenting and adolescent self regulation. Journal of Child Psychology and Psychiatry, 52(5), 619-626. ‐ ‐ Benton, D. (2007). The impact of diet on anti-social, violent and criminal behaviour. Neuroscience & Biobehavioral Reviews, 31(5), 752-774.

Benton, D. (2008). The influence of children’s diet on their cognition and behavior. European Journal of Nutrition, 47(3), 25-37.

Bhutta, A. T., Cleves, M. A., Casey, P. H., Cradock, M. M., & Anand, K. J. S. (2002). Cognitive and behavioral outcomes of school-aged children who were born preterm: A meta- analysis. Jama, 288(6), 728-737.

Bigelow, A. E., Power, M., Gillis, D. E., Maclellan Peters, J., Alex, M., & McDonald, C. (2014). Breastfeeding, skin-to-skin contact, and mother-infant interactions over infants’ first three months. Infant Mental Health Journal‐, 35(1), 51-62.

Black, M. M. (2008). Effects of vitamin B12 and folate deficiency on brain development in children. Food and Nutrition Bulletin, 29(2 Suppl), S126.

Blokland, A. A. J., Nagin, D., & Nieuwbeerta, P. (2005). Life span offending trajectories of a Dutch conviction cohort. Criminology, 43, 919.

Boersma, M., Smit, D. J., Boomsma, D. I., De Geus, E. J., Delemarre-van de Waal, H. A., & Stam, C. J. (2013). Growing trees in child brains: Graph theoretical analysis of electroencephalography-derived minimum spanning tree in 5-and 7-year-old children reflects brain maturation. Brain Connectivity, 3(1), 50-60.

Boldt, L. J., Kochanska, G., Yoon, J. E., & Koenig Nordling, J. (2014). Children’s attachment to both parents from toddler age to middle childhood: Links to adaptive and maladaptive outcomes. Attachment & Human Development, 16(3), 211-229.

Bonta, J., & Andrews, D. A. (2007). Risk-need-responsivity model for offender assessment and rehabilitation. Rehabilitation, 6, 1-22.

Borra, C., Iacovou, M., & Sevilla, A. (2012). The effect of breastfeeding on children's cognitive and noncognitive development. Labour Economics, 19(4), 496-515.

265 Bost, K. K., Wiley, A. R., Fiese, B., Hammons, A., McBride, B., & STRONG KIDS Team. (2014). Associations between adult attachment style, emotion regulation, and preschool children's food consumption. Journal of Developmental & Behavioral Pediatrics, 35(1), 50-61.

Botting, N., Powls, A., Cooke, R. W., & Marlow, N. (1997). Attention deficit hyperactivity disorders and other psychiatric outcomes in very low birthweight children at 12 years. Journal of Child Psychology and Psychiatry, 38(8), 931-941.

Boutet, C., Vercueil, L., Schelstraete, C., Buffin, A., & Legros, J. J. (2006, June). Oxytocin and maternal stress during the post-partum period. Annales D'endocrinologie, 67(3), 214.

Boutwell, B. B., & Beaver, K. M. (2010). Maternal cigarette smoking during pregnancy and offspring externalizing behavioral problems: A propensity score matching analysis International Journal of Environmental Research and Public Health, 7(1), 146-163.

Boutwell, B. B., Beaver, K. M., & Barnes, J. C. (2012). Role of breastfeeding in childhood cognitive development: A propensity score matching analysis. Journal of Paediatrics and Child Health, 48(9), 840-845.

Bowlby, J. (1969). Attachment. Attachment and Loss: Vol. 1. Loss. New York: Basic Books.

Bowlby, J. (1982). Attachment and loss: Retrospect and prospect. American Journal of Orthopsychiatry, 52(4), 664.

Bozorgi, Z. D., Sho’leh Amiri, A. M., & Talebi, H (2014). Effectiveness of attachment-based intervention in eating behaviors of obese girls. International Journal of Psychology and Behavioral Research, 3(6), 572-580.

Braet, C., & van Strien, T. (1997). Assessment of emotional, externally induced and restrained eating behaviour in nine to twelve-year-old obese and non-obese children. Behaviour Research and Therapy, 35(9), 863-873.

Brame, R., Bushway, S. D., & Paternoster, R. (2003). Examining the prevalence of criminal desistance. Criminology, 41(2), 423-448.

Brandt, K. A., Andrews, C. M., & Kvale, J. (1998). Mother infant interaction and breastfeeding outcome 6 weeks after birth. Journal of Obstetric, Gynecologic, & Neonatal Nursing, 27(2), 169-174. ‐

Brendgen, M., Dionne, G., Girard, A., Boivin, M., Vitaro, F., & Pérusse, D. (2005). Examining genetic and environmental effects on social aggression: A study of 6 year old twins. Child Development, 76(4), 930-946. ‐ ‐

266 Breslau, N., Brown, G. G., DelDotto, J. E., Kumar, S., Ezhuthachan, S., Andreski, P., & Hufnagle, K. G. (1996). Psychiatric sequelae of low birth weight at 6 years of age. Journal of Abnormal Child Psychology, 24(3), 385-400.

Breslau, N., & Chilcoat, H. D. (2000). Psychiatric sequelae of low birth weight at 11 years of age. Biological Psychiatry, 47(11), 1005-1011.

Breslau, N., Chilcoat, H., DelDotto, J., Andreski, P., & Brown, G. (1996). Low birth weight and neurocognitive status at six years of age. Biological Psychiatry, 40(5), 389-397.

Brion, M. J., Lawlor, D. A., Matijasevich, A., Horta, B., Anselmi, L., Araújo, C. L., ... & Smith, G. D. (2011). What are the causal effects of breastfeeding on IQ, obesity and blood pressure? Evidence from comparing high-income with middle-income cohorts. International Journal of Epidemiology, 40(3), 670-680.

Britton, J. R., Britton, H. L., & Gronwaldt, V. (2006). Breastfeeding, sensitivity, and attachment. Pediatrics, 118(5), e1436-e1443.

Broidy, L. M., Nagin, D. S., Tremblay, R. E., Bates, J. E., Brame, B., Dodge, K. A., ... & Vitaro, F. (2003). Developmental trajectories of childhood disruptive behaviors and adolescent delinquency: A six-site, cross-national study. Developmental Psychology, 39(2), 222.

Bronte Tinkew, J., Scott, M. E., & Horowitz, A. (2009). Male pregnancy intendedness and children's mental proficiency and attachment security during toddlerhood. Journal of ‐ Marriage and Family, 71(4), 1001-1025.

Brown, T. T., & Jernigan, T. L. (2012). Brain development during the preschool years. Neuropsychology Review, 22(4), 313-333.

Brumariu, L. E., Kerns, K. A., & Seibert, A. (2012). Mother–child attachment, emotion regulation, and anxiety symptoms in middle childhood. Personal Relationships, 19(3), 569-585.

Buschgens, C. J., Swinkels, S. H., van Aken, M. A., Ormel, J., Verhulst, F. C., & Buitelaar, J. K. (2009). Externalizing behaviors in preadolescents: Familial risk to externalizing behaviors, prenatal and perinatal risks, and their interactions. European Child & Adolescent Psychiatry, 18(2), 65-74.

Calkins, S. D., Gill, K. L., Johnson, M. C., & Smith, C. L. (1999). Emotional reactivity and emotional regulation strategies as predictors of social behavior with peers during toddlerhood. Social Development, 8(3), 310-334.

Calkins, S. D., & Leerkes, E. M. (2004). Early attachment processes and the development of emotional self-regulation. In R. F. Baumeister & K. D. Vohs (Eds.), Handbook of self- regulation: Research, theory, and applications (pp. 324-339). New York: The Guilford Press.

267 Campbell, S. B., Shaw, D. S., & Gilliom, M. (2000). Early externalizing behavior problems: Toddlers and preschoolers at risk for later maladjustment. Development and Psychopathology, 12(03), 467-488.

Casey, P. H., Szeto, K., Lensing, S., Bogle, M., & Weber, J. (2001). Children in food- insufficient, low-income families: Prevalence, health, and nutrition status. Archives of Pediatrics & Adolescent Medicine, 155(4), 508-514.

Caspi, A., Henry, B., McGee, R. O., Moffitt, T. E., & Silva, P. A. (1995). Temperamental origins of child and adolescent behavior problems: From age three to age fifteen. Child Development, 66(1), 55-68.

Caspi, A., McClay, J., Moffitt, T. E., Mill, J., Martin, J., Craig, I. W., ... & Poulton, R. (2002). Role of genotype in the cycle of violence in maltreated children. Science, 297(5582), 851-854.

Caspi, A., Moffitt, T. E., Newman, D. L., & Silva, P. A. (1996). Behavioral observations at age 3 years predict adult psychiatric disorders: Longitudinal evidence from a birth cohort. Archives of General Psychiatry, 53(11), 1033-1039.

Caspi, A., Williams, B., Kim-Cohen, J., Craig, I. W., Milne, B. J., Poulton, R., ... & Moffitt, T. E. (2007). Moderation of breastfeeding effects on the IQ by genetic variation in fatty acid metabolism. Proceedings of the National Academy of Sciences, 104(47), 18860-18865.

Castelao, C. F., & Kröner-Herwig, B. (2014). Developmental trajectories and predictors of externalizing behavior: A comparison of girls and boys. Journal of Youth and Adolescence, 43(5), 775-789.

Castellote, C., Casillas, R., Ramírez-Santana, C., Pérez-Cano, F. J., Castell, M., Moretones, M. G., ... & Franch, À. (2011). Premature delivery influences the immunological composition of colostrum and transitional and mature human milk. The Journal of Nutrition, 141(6), 1181-1187.

Cauffman, E., Steinberg, L., & Piquero, A. R. (2005). Psychological, neuropsychological, and physiological correlates of serious antisocial behavior in adolescence: The role of self- control. Criminology, 43(1), 133-176.

Choe, D. E., Olson, S. L., & Sameroff, A. J. (2013). Effects of early maternal distress and parenting on the development of children's self-regulation and externalizing behavior. Development and Psychopathology, 25(02), 437-453.

Clark, C. A., Woodward, L. J., Horwood, L. J., & Moor, S. (2008). Development of emotional and behavioral regulation in children born extremely preterm and very preterm: Biological and social influences. Child Development, 79(5), 1444-1462.

268 Colen, C. G., & Ramey, D. M. (2014). Is breast truly best? Estimating the effects of breastfeeding on long-term child health and wellbeing in the United States using sibling comparisons. Social Science & Medicine, 109, 55-65.

Cohen, D. J., Dibble, E., Grawe, J. M., & Pollin, W. (1975). Reliably separating identical from fraternal twins. Archives of General Psychiatry, 32(11), 1371-1375.

Colodro-Conde, L., Sánchez-Romera, J. F., & Ordoñana, J. R. (2013). Heritability of initiation and duration of breastfeeding behavior. Twin Research and Human Genetics, 16(02), 575-580.

Colodro-Conde, L., Zhu, G., Power, R. A., Henders, A., Heath, A. C., Madden, P. A., ... & Martin, N. G. (2014). A twin study of breastfeeding with a preliminary genome-wide association scan. Twin Research and Human Genetics, 1-12.

Colonnesi, C., Draijer, E. M., Jan JM Stams, G., Van der Bruggen, C. O., Bögels, S. M., & Noom, M. J. (2011). The relation between insecure attachment and child anxiety: A meta-analytic review. Journal of Clinical Child & Adolescent Psychology, 40(4), 630- 645.

Connell, A. M., & Goodman, S. H. (2002). The association between psychopathology in fathers versus mothers and children's internalizing and externalizing behavior problems: A meta- analysis. Psychological Bulletin, 128(5), 746.

Cooke, R. W. (2006). Are there critical periods for brain growth in children born preterm?. Archives of Disease in Childhood-Fetal and Neonatal Edition, 91(1), F17-F20.

Corapci, F., Radan, A. E., & Lozoff, B. (2006). Iron deficiency in infancy and mother-child interaction at 5 years. Journal of Developmental and Behavioral Pediatrics, 27(5), 371.

Cornelis, M. C., & Hu, F. B. (2012). Gene-environment interactions in the development of type 2 diabetes: Recent progress and continuing challenges. Annual Review of Nutrition, 32, 245-259.

Coyl, D. D., Roggman, L. A., & Newland, L. A. (2002). Stress, maternal depression, and negative mother–infant interactions in relation to infant attachment. Infant Mental Health Journal, 23(1 2), 145-163.

Cronk, N. J., Slutske,‐ W. S., Madden, P. A., Bucholz, K. K., Reich, W., & Heath, A. C. (2002). Emotional and behavioral problems among female twins: An evaluation of the equal environments assumption. Journal of the American Academy of Child & Adolescent Psychiatry, 41(7), 829-837.

Dauncey, M. J. (2012). Recent advances in nutrition, genes and brain health. Proceedings of the Nutrition Society, 71(04), 581-591.

269 Deater-Deckard, K., Dodge, K. A., Bates, J. E., & Pettit, G. S. (1997). Multiple risk factors in the development of externalizing behavior problems: Group and individual differences. Development and Psychopathology, 10(3), 469-493.

Deater Deckard, K., & Plomin, R. (1999). An adoption study of the etiology of teacher and parent reports of externalizing behavior problems in middle childhood. Child ‐ Development, 70(1), 144-154.

DeFries, J. C., & Fulker, D. W. (1985). Multiple regression analysis of twin data. Behavior Genetics, 15(5), 467-473.

DeFries, J. C., & Fulker, D. W. (1988). Multiple regression analysis of twin data: Etiology of deviant scores versus individual differences. Acta Geneticae Medicae et Gemellologiae: Twin Research, 37(3-4), 205-216.

Dieterich, C. M., Felice, J. P., O’Sullivan, E., & Rasmussen, K. M. (2013). Breastfeeding and health outcomes for the mother-infant dyad. Pediatric Clinics of North America, 60(1), 31-48.

Delobel-Ayoub, M., Kaminski, M., Marret, S., Burguet, A., Marchand, L., Sylvie, N., ... & Larroque, B. (2006). Behavioral outcome at 3 years of age in very preterm infants: The EPIPAGE study. Pediatrics, 117(6), 1996-2005.

Deoni, S. C., Dean III, D. C., Piryatinsky, I., O'Muircheartaigh, J., Waskiewicz, N., Lehman, K., ... & Dirks, H. (2013). Breastfeeding and early white matter development: A cross- sectional study. NeuroImage, 82, 77-86.

Diener, M. L., Mangelsdorf, S. C., McHale, J. L., & Frosch, C. A. (2002). Infants' behavioral strategies for emotion regulation with fathers and mothers: Associations with emotional expressions and attachment quality. Infancy, 3(2), 153-174.

Dewey, K. G. (2003). Is breastfeeding protective against child obesity?. Journal of Human Lactation, 19(1), 9-18.

Dodge, K. A., & Pettit, G. S. (2003). A biopsychosocial model of the development of chronic conduct problems in adolescence. Developmental Psychology, 39(2), 349-371.

Dodge, K. A., Pettit, G. S., & Bates, J. E. (1994). Socialization mediators of the relation between socioeconomic status and child conduct problems. Child Development, 65(2), 649-665.

D'Onofrio, B. M., Hulle, C. A. V., Waldman, I. D., Rodgers, J. L., Harden, K. P., Rathouz, P. J., & Lahey, B. B. (2008). Smoking during pregnancy and offspring externalizing problems: An exploration of genetic and environmental confounds. Development and Psychopathology, 20(01), 139-164.

270 D’Onofrio, B. M., Lahey, B. B., Turkheimer, E., & Lichtenstein, P. (2013). Critical need for family-based, quasi-experimental designs in integrating genetic and social science research. American Journal of Public Health, 103(S1), S46-S55.

Dubois-Comtois, K., Moss, E., Cyr, C., & Pascuzzo, K. (2013). Behavior problems in middle childhood: The predictive role of maternal distress, child attachment, and mother-child interactions. Journal of Abnormal Child Psychology, 41(8), 1311-1324.

Duijts, L., Jaddoe, V. W., Hofman, A., & Moll, H. A. (2010). Prolonged and exclusive breastfeeding reduces the risk of infectious diseases in infancy. Pediatrics, 126(1), e18- e25.

Dvorak, B., Fituch, C. C., Williams, C. S., Hurst, N. M., & Schanler, R. J. (2003). Increased epidermal growth factor levels in human milk of mothers with extremely premature infants. Pediatric Research, 54(1), 15-19.

Ekström, A., & Nissen, E. (2006). A mother's feelings for her infant are strengthened by excellent breastfeeding counseling and continuity of care. Pediatrics, 118(2), e309-e314.

Elgen, I., Sommerfelt, K., & Markestad, T. (2002). Population based, controlled study of behavioural problems and psychiatric disorders in low birthweight children at 11 years of age. Archives of Disease in Childhood-Fetal and Neonatal Edition, 87(2), F128-F132.

Else-Quest, N. M., Hyde, J. S., & Clark, R. (2003). Breastfeeding, bonding, and the mother- infant relationship. Merrill-Palmer Quarterly (1982-), 495-517.

Espy, K. A., Sheffield, T. D., Wiebe, S. A., Clark, C. A., & Moehr, M. J. (2011). Executive control and dimensions of problem behaviors in preschool children. Journal of Child Psychology and Psychiatry, 52(1), 33-46.

Evenhouse, E., & Reilly, S. (2005). Improved estimates of the benefits of breastfeeding using sibling comparisons to reduce selection bias. Health Services Research, 40(6p1), 1781- 1802.

Faber, A., & Dubé, L. (2015). Parental attachment insecurity predicts child and adult high- caloric food consumption. Journal of Health Psychology, 20(5), 511-524.

Faith, M. S., Berkowitz, R. I., Stallings, V. A., Kerns, J., Storey, M., & Stunkard, A. J. (2006). Eating in the absence of hunger: A genetic marker for childhood obesity in prepubertal boys?. Obesity, 14(1), 131-138.

Farrington, D. P. (2003). Developmental and life-course criminology: Key theoretical and empirical issues-the 2002 Sutherland award address. Criminology, 41, 221.

271 Fearon, R. P., Bakermans Kranenburg, M. J., Van IJzendoorn, M. H., Lapsley, A. M., & Roisman, G. I. (2010). The significance of insecure attachment and disorganization in the development of chil‐ dren’s externalizing behavior: A meta analytic study. Child Development, 81(2), 435-456. ‐ Feldman, R., & Eidelman, A. I. (2003). Direct and indirect effects of breast milk on the neurobehavioral and cognitive development of premature infants. Developmental Psychobiology, 43(2), 109-119.

Feldman, R., Gordon, I., & Zagoory Sharon, O. (2011). Maternal and paternal plasma, salivary, and urinary oxytocin and parent–infant synchrony: Considering stress and affiliation components of human bonding.‐ Developmental Science, 14(4), 752-761.

Felson, J. (2014). What can we learn from twin studies? A comprehensive evaluation of the equal environments assumption. Social Science Research, 43, 184-199.

Ferguson, C. J. (2010). Genetic contributions to antisocial personality and behavior: A meta- analytic review from an evolutionary perspective. The Journal of Social Psychology, 150(2), 160-180.

Fergusson, D. M., Boden, J. M., & Horwood, L. J. (2014). Bullying in childhood, externalizing behaviors, and adult offending: Evidence from a 30-year study. Journal of School Violence, 13(1), 146-164.

Fergusson, D. M., & Horwood, L. J. (1995). Early disruptive behavior, IQ, and later school achievement and delinquent behavior. Journal of Abnormal Child Psychology, 23(2), 183-199.

Fewtrell, M., Wilson, D. C., Booth, I., & Lucas, A. (2011). Six months of exclusive breast feeding: How good is the evidence?. BMJ, 342.

Field, S. S. (2014). Interaction of genes and nutritional factors in the etiology of autism and attention deficit/hyperactivity disorders: A case control study. Medical Hypotheses, 82(6), 654-661.

Flacking, R., Nyqvist, K. H., Ewald, U., & Wallin, L. (2003). Long-term duration of breastfeeding in Swedish low birth weight infants. Journal of Human Lactation, 19(2), 157-165.

Fox, S. E., Levitt, P., & Nelson III, C. A. (2010). How the timing and quality of early experiences influence the development of brain architecture. Child Development, 81(1), 28-40.

Frosch, C. A., Mangelsdorf, S. C., & McHale, J. L. (2000). Marital behavior and the security of preschooler–parent attachment relationships. Journal of Family Psychology, 14(1), 144.

272 Galler, J. R., Bryce, C. P., Waber, D. P., Hock, R. S., Harrison, R., Eaglesfield, G. D., & Fitzmaurice, G. (2012). Infant malnutrition predicts conduct problems in adolescents. Nutritional Neuroscience, 15(4), 186-192.

Galler, J. R., Bryce, C. P., Waber, D. P., Medford, G., Eaglesfield, G. D., & Fitzmaurice, G. (2011). Early malnutrition predicts parent reports of externalizing behaviors at ages 9– 17. Nutritional Neuroscience, 14(4), 138-144.

Garver, W. S., Newman, S. B., Gonzales-Pacheco, D. M., Castillo, J. J., Jelinek, D., Heidenreich, R. A., & Orlando, R. A. (2013). The genetics of childhood obesity and interaction with dietary macronutrients. Genes & Nutrition, 8(3), 271-287.

Georgieff, M. K. (2007). Nutrition and the developing brain: Nutrient priorities and measurement. The American Journal of Clinical Nutrition, 85(2), 614S-620S.

Gershoff, E. T. (2002). Corporal punishment by parents and associated child behaviors and experiences: A meta-analytic and theoretical review. Psychological Bulletin, 128(4), 539.

Gesch, C. B., Hammond, S. M., Hampson, S. E., Eves, A., & Crowder, M. J. (2002). Influence of supplementary vitamins, minerals and essential fatty acids on the antisocial behaviour of young adult prisoners: Randomised, placebo-controlled trial. The British Journal of Psychiatry, 181(1), 22-28.

Ghods, E., Kreissl, A., Brandstetter, S., Fuiko, R., & Widhalm, K. (2011). Head circumference catch-up growth among preterm very low birth weight infants: Effect on neurodevelopmental outcome. Journal of Perinatal Medicine, 39(5), 579-586.

Gini, G., Pozzoli, T., & Hymel, S. (2014). Moral disengagement among children and youth: A meta analytic review of links to aggressive behavior. Aggressive Behavior, 40(1), 56-68.

Golding, J., Emmett,‐ P., Iles-Caven, Y., Steer, C., & Lingam, R. (2014). A review of environmental contributions to childhood motor skills. Journal of Child Neurology, 29(11), 1531-1547.

Goldner, L., & Scharf, M. (2013). Attachment security, personality, and adjustment of elementary school children. The Journal of Genetic Psychology, 174(5), 473-493.

Goldsmith, H. (1991). A zygosity questionnaire for young twins: A research note. Behavior Genetics, 21(3), 257-269.

Golub, M., & Hogrefe, C. (2014). Prenatal iron deficiency and monoamine oxidase A (MAOA) polymorphisms: Combined risk for later cognitive performance in rhesus monkeys. Genes & Nutrition, 9(2), 1-9.

Gómez-Pinilla, F. (2008). Brain foods: the effects of nutrients on brain function. Nature Reviews Neuroscience, 9(7), 568-578.

273 Goossens, L., Braet, C., Bosmans, G., & Decaluwé, V. (2011). Loss of control over eating in pre- adolescent youth: The role of attachment and self-esteem. Eating Behaviors, 12(4), 289- 295.

Gottfredson, M. R., & Hirschi, T. (1990). A general theory of crime. Stanford University Press. Grantham-McGregor, S., & Ani, C. (2001). A review of studies on the effect of iron deficiency on cognitive development in children. The Journal of Nutrition, 131(2), 649S- 668S.

Gow, R. V., Vallee-Tourangeau, F., Crawford, M. A., Taylor, E., Ghebremeskel, K., Bueno, A. A., ... & Rubia, K. (2013). Omega-3 fatty acids are inversely related to callous and unemotional traits in adolescent boys with attention deficit hyperactivity disorder. Prostaglandins, Leukotrienes and Essential Fatty Acids (PLEFA), 88(6), 411-418.

Green, A. E., Gesten, E. L., Greenwald, M. A., & Salcedo, O. (2008). Predicting delinquency in adolescence and young adulthood: A longitudinal analysis of early risk factors. Youth Violence and Juvenile Justice, 6(4), 323-342.

Greenough, W. T., Black, J. E., & Wallace, C. S. (1987). Experience and brain development. Child Development, 539-559.

Gribble, K. D. (2006). Mental health, attachment and breastfeeding: Implications for adopted children and their mothers. International Breastfeeding Journal, 1, 5-15.

Grieger, J. A., Scott, J., & Cobiac, L. (2011). Dietary patterns and breast-feeding in Australian children. Public Health Nutrition, 14(11), 1939-1947.

Groen Blokhuis, M. M., Franić, S., van Beijsterveldt, C. E., de Geus, E., Bartels, M., Davies, G. E., ... & Boomsma, D. I. (2013). A prospective study of the effects of breastfeeding and ‐ FADS2 polymorphisms on cognition and hyperactivity/attention problems. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics, 162(5), 457-465.

Groh, A. M., Roisman, G. I., van IJzendoorn, M. H., Bakermans Kranenburg, M. J., & Fearon, R. (2012). The significance of insecure and disorganized attachment for children’s internalizing symptoms: A meta analytic study. Child Development‐ , 83(2), 591-610.

Guttmann-Steinmetz, S., & Crowell, J. A.‐ (2006). Attachment and externalizing disorders: A developmental psychopathology perspective. Journal of the American Academy of Child & Adolescent Psychiatry, 45(4), 440-451.

Hack, M., Taylor, H. G., Schluchter, M., Andreias, L., Drotar, D., & Klein, N. (2009). Behavioral outcomes of extremely low birth weight children at age 8 years. Journal of Developmental & Behavioral Pediatrics, 30(2), 122-130.

Hanson, L. A. (1998). Breastfeeding provides passive and likely long-lasting active immunity. Annals of Allergy, Asthma & Immunology, 81(6), 523-537.

274 Hanson, J. L., Hair, N., Shen, D. G., Shi, F., Gilmore, J. H., Wolfe, B. L., & Pollak, S. D. (2013). Family poverty affects the rate of human infant brain growth. PloS one, 8(12), e80954.

Hanson, N. I., Neumark-Sztainer, D., Eisenberg, M. E., Story, M., & Wall, M. (2005). Associations between parental report of the home food environment and adolescent intakes of fruits, vegetables and dairy foods. Public Health Nutrition, 8(01), 77-85.

Harder, T., Bergmann, R., Kallischnigg, G., & Plagemann, A. (2005). Duration of breastfeeding and risk of overweight: A meta-analysis. American Journal of Epidemiology, 162(5), 397-403.

Hayatbakhsh, M. R., O'Callaghan, M. J., Bor, W., Williams, G. M., & Najman, J. M. (2012). Association of breastfeeding and adolescents' psychopathology: A large prospective study. Breastfeeding Medicine, 7(6), 480-486.

Haynie, D. L., & McHugh, S. (2003). Sibling deviance: In the shadows of mutual and unique friendship effects?. Criminology, 41(2), 355-392.

Heikkilä, K., Sacker, A., Kelly, Y., Renfrew, M. J., & Quigley, M. A. (2011). Breast feeding and child behaviour in the Millennium Cohort Study. Archives of Disease in Childhood, 96(7), 635-642.

Helpern, J. A., Adisetiyo, V., Falangola, M. F., Hu, C., Di Martino, A., Williams, K., ... & Jensen, J. H. (2011). Preliminary evidence of altered gray and white matter microstructural development in the frontal lobe of adolescents with attention deficit hyperactivity disorder: A diffusional kurtosis imaging study. Journal of Magnetic Resonance Imaging, 33(1), 17-23. ‐

Henry, B., Caspi, A., Moffitt, T. E., & Silva, P. A. (1996). Temperamental and familial predictors of violent and nonviolent criminal convictions: Age 3 to age 18. Developmental Psychology, 32(4), 614.

Hernell, O., Bläckberg, L., Chen, Q., Sternby, B., & Nilsson, Å. (1993). Does the bile salt- stimulated lipase of human milk have a role in the use of the milk long-chain polyunsaturated fatty acids?. Journal of Pediatric Gastroenterology and Nutrition, 16(4), 426-431.

Hibbeln, J. R., Ferguson, T. A., & Blasbalg, T. L. (2006). Omega-3 fatty acid deficiencies in neurodevelopment, aggression and autonomic dysregulation: Opportunities for intervention. International Review of Psychiatry, 18(2), 107-118.

Horwood, L. J., Darlow, B. A., & Mogridge, N. (2001). Breast milk feeding and cognitive ability at 7–8 years. Archives of Disease in Childhood-Fetal and Neonatal Edition, 84(1), F23- F27.

275 Hosaka, M., Asayama, K., Staessen, J. A., Ohkubo, T., Hayashi, K., Tatsuta, N., ... & Satoh, H. (2012). Breastfeeding leads to lower blood pressure in 7-year-old Japanese children: Tohoku Study of Child Development. Hypertension Research, 36(2), 117-122.

Howard, A. L., Robinson, M., Smith, G. J., Ambrosini, G. L., Piek, J. P., & Oddy, W. H. (2011). ADHD is associated with a “Western” dietary pattern in adolescents. Journal of Attention Disorders, 15(5), 403-411.

Huhtala, M., Korja, R., Lehtonen, L., Haataja, L., Lapinleimu, H., & Rautava, P. (2012). Parental psychological well-being and behavioral outcome of very low birth weight infants at 3 years. Pediatrics, 129(4), e937-e944.

Hultman, C. M., Torrång, A., Tuvblad, C., Cnattingius, S., Larsson, J. O., & Lichtenstein, P. (2007). Birth weight and attention-deficit/hyperactivity symptoms in childhood and early adolescence: A prospective Swedish twin study. Journal of the American Academy of Child & Adolescent Psychiatry, 46(3), 370-377.

Huntsinger, E. T., & Luecken, L. J. (2004). Attachment relationships and health behavior: The mediational role of self-esteem. Psychology & Health, 19(4), 515-526.

Hyatt, C. J., Haney-Caron, E., & Stevens, M. C. (2012). Cortical thickness and folding deficits in conduct-disordered adolescents. Biological Psychiatry, 72(3), 207-214.

Innis, S. M. (2007). Dietary (n-3) fatty acids and brain development. The Journal of Nutrition, 137(4), 855-859.

Isaacs, E. B., Fischl, B. R., Quinn, B. T., Chong, W. K., Gadian, D. G., & Lucas, A. (2010). Impact of breast milk on intelligence quotient, brain size, and white matter development. Pediatric Research, 67(4), 357-362.

Jaccard, J., Wan, C. K., & Turrisi, R. (1990). The detection and interpretation of interaction effects between continuous variables in multiple regression. Multivariate Behavioral Research, 25(4), 467-478.

Jackson, D. B., & Beaver, K. M. (2013). The influence of neuropsychological deficits in early childhood on low self-control and misconduct through early adolescence. Journal of Criminal Justice, 41(4), 243-251.

Jackson, D.B., & Beaver, K. M. (2015). Sibling differences in low birth weight, dopaminergic polymorphisms, and ADHD symptomatology: Evidence of GxE. Psychiatry Research, 226(2-3), 467-473.

Jackson, D. B., & Beaver, K. M. (in press). The association between breastfeeding exposure and duration, neuropsychological deficits, and psychopathic personality traits in offspring: The moderating role of 5HTTLPR. Psychiatric Quarterly.

276 Jackson, K. M., & Nazar, A. M. (2006). Breastfeeding, the immune response, and long-term health. JAOA: Journal of the American Osteopathic Association, 106(4), 203-207.

Jacobsen, T., Edelstein, W., & Hofmann, V. (1994). A longitudinal study of the relation between representations of attachment in childhood and cognitive functioning in childhood and adolescence. Developmental Psychology, 30(1), 112.

Jaekel, J., Pluess, M., Belsky, J., & Wolke, D. (2014). Effects of maternal sensitivity on low birth weight children's academic achievement: A test of differential susceptibility versus diathesis stress. Journal of Child Psychology and Psychiatry. DOI: 10.1111/jcpp.12331

Jaffee, S. R., Caspi, A., Moffitt, T. E., Dodge, K. A., Rutter, M., Taylor, A., & Tully, L. A. (2005). Nature× nurture: Genetic vulnerabilities interact with physical maltreatment to promote conduct problems. Development and Psychopathology, 17(01), 67-84.

Jaffee, S. R., Strait, L. B., & Odgers, C. L. (2012). From correlates to causes: can quasi- experimental studies and statistical innovations bring us closer to identifying the causes of antisocial behavior?. Psychological Bulletin, 138(2), 272.

Jansen, J., Weerth, C. D., & Riksen-Walraven, J. M. (2008). Breastfeeding and the mother–infant relationship—a review. Developmental Review, 28(4), 503-521.

Jennings, W. G., & Reingle, J. M. (2012). On the number and shape of developmental/life-course violence, aggression, and delinquency trajectories: A state-of-the-art review. Journal of Criminal Justice, 40(6), 472-489.

Jones, C. (2013). The road to maternal responsiveness is paved with good intentions: an investigation into the relative effects of breastfeeding intention and practice on observed maternal responsiveness after birth (Doctoral dissertation, University of Exeter).

Julvez, J., Guxens, M., Carsin, A. E., Forns, J., Mendez, M., Turner, M. C., & Sunyer, J. (2014). A cohort study on full breastfeeding and child neuropsychological development: The role of maternal social, psychological, and nutritional factors. Developmental Medicine & Child Neurology, 56(2), 148-156.

Julvez, J., Ribas Fitó, N., Forns, M., Garcia Esteban, R., Torrent, M., & Sunyer, J. (2007). Attention behaviour and hyperactivity at age 4 and duration of breast feeding. Acta Paediatrica‐ , 96(6), 842-847. ‐ ‐ Kafouri, S., Kramer, M., Leonard, G., Perron, M., Pike, B., Richer, L., ... & Paus, T. (2012). Breastfeeding and brain structure in adolescence. International Journal of Epidemiology, 42(1), 150-159.

Kamkar, K., Doyle, A. B., & Markiewicz, D. (2012). Insecure attachment to parents and depressive symptoms in early adolescence: Mediating roles of attributions and self- esteem. International Journal of Psychological Studies, 4(2), 3-18.

277 Kandel, E., & Mednick, S. A. (1991). Perinatal complications predict violent offending. Criminology, 29(3), 519-529.

Kaplan, B. J., Fisher, J. E., Crawford, S. G., Field, C. J., & Kolb, B. (2004). Improved mood and behavior during treatment with a mineral-vitamin supplement: An open-label case series of children. Journal of Child and Adolescent Psychopharmacology, 14(1), 115-122.

Karnehed, N., Tynelius, P., Heitman, B. L., & Rasmussen, F. (2006). Physical activity, diet and gene–environment interactions in relation to body mass index and waist circumference: The Swedish young male twins study. Public Health Nutrition, 9(07), 851-858.

Katz, D. L., Njike, V. Y., Rhee, L. Q., Reingold, A., & Ayoob, K. T. (2010). Performance characteristics of NuVal and the Overall Nutritional Quality Index (ONQI). The American Journal of Clinical Nutrition, 91(4), 1102S-1108S.

Keiley, M. K., Bates, J. E., Dodge, K. A., & Pettit, G. S. (2000). A cross-domain growth analysis: Externalizing and internalizing behaviors during 8 years of childhood. Journal of Abnormal Child Psychology, 28(2), 161-179.

Kemp, L., Harris, E., McMahon, C., Matthey, S., Vimpani, G., Anderson, T., ... & Zapart, S. (2011). Child and family outcomes of a long-term nurse home visitation programme: A randomised controlled trial. Archives of Disease in Childhood, 96(6), 533-540.

Kendler, K. S., Neale, M. C., Kessler, R. C., Heath, A. C., & Eaves, L. J. (1993). A test of the equal-environment assumption in twin studies of psychiatric illness. Behavior Genetics, 23(1), 21-27.

Kerns, K. A., Brumariu, L. E., & Seibert, A. (2011). Multi-method assessment of mother-child attachment: Links to parenting and child depressive symptoms in middle childhood. Attachment & Human Development, 13(4), 315-333.

Kerr, D. C., Lopez, N. L., Olson, S. L., & Sameroff, A. J. (2004). Parental discipline and externalizing behavior problems in early childhood: The roles of moral regulation and child gender. Journal of Abnormal Child Psychology, 32(4), 369-383.

Kim, P., Feldman, R., Mayes, L. C., Eicher, V., Thompson, N., Leckman, J. F., & Swain, J. E. (2011). Breastfeeding, brain activation to own infant cry, and maternal sensitivity. Journal of Child Psychology and Psychiatry, 52(8), 907-915.

Kitsao-Wekulo, P. K., Holding, P. A., Taylor, H. G., Kvalsvig, J. D., & Connolly, K. J. (2013). Determinants of variability in motor performance in middle childhood: A cross-sectional study of balance and motor co-ordination skills. BMC Psychology, 1(1), 29.

Knopik, V. S. (2009). Maternal smoking during pregnancy and child outcomes: Real or spurious effect?. Developmental Neuropsychology, 34(1), 1-36.

278 Knopik, V. S., Heath, A. C., Jacob, T., Slutske, W. S., Bucholz, K. K., Madden, P. A., ... & Martin, N. G. (2006). Maternal alcohol use disorder and offspring ADHD: Disentangling genetic and environmental effects using a children-of-twins design. Psychological Medicine, 36(10), 1461-1471.

Kochanska, G., & Kim, S. (2013). Early attachment organization with both parents and future behavior problems: From infancy to middle childhood. Child Development, 84(1), 283- 296.

Kohler, H. P., & Rodgers, J. L. (2001). DF-analyses of heritability with double-entry twin data: Asymptotic standard errors and efficient estimation. Behavior Genetics, 31(2), 179-191.

Kokko, K., Tremblay, R. E., Lacourse, E., Nagin, D. S., & Vitaro, F. (2006). Trajectories of prosocial behavior and physical aggression in middle childhood: Links to adolescent school dropout and physical violence. Journal of Research on Adolescence, 16(3), 403- 428.

Kolb, B., & Gibb, R. (2011). Brain plasticity and behaviour in the developing brain. Journal of the Canadian Academy of Child and Adolescent Psychiatry, 20(4), 265.

Kramer, M. S., Aboud, F., Mironova, E., Vanilovich, I., Platt, R. W., Matush, L., ... & Shapiro, S. (2008a). Breastfeeding and child cognitive development: New evidence from a large randomized trial. Archives of General Psychiatry,65(5), 578-584.

Kramer, M. S., Fombonne, E., Igumnov, S., Vanilovich, I., Matush, L., Mironova, E., ... & Platt, R. W. (2008b). Effects of prolonged and exclusive breastfeeding on child behavior and maternal adjustment: Evidence from a large, randomized trial. Pediatrics, 121(3), e435- e440.

Kramer, M. S., Fombonne, E., Matush, L., Bogdanovich, N., Dahhou, M., & Platt, R. W. (2011). Long term behavioural consequences of infant feeding: The limits of observational studies. Paediatric and Perinatal Epidemiology, 25(6), 500-506. ‐ Kramer, M. S., Guo, T., Platt, R. W., Sevkovskaya, Z., Dzikovich, I., Collet, J. P., ... & Bogdanovich, N. (2003). Infant growth and health outcomes associated with 3 compared with 6 mo of exclusive breastfeeding. The American Journal of Clinical Nutrition, 78(2), 291-295.

Kramer, M. S., Matush, L., Vanilovich, I., Platt, R. W., Bogdanovich, N., Sevkovskaya, Z., ... & Shapiro, S. (2007). Effects of prolonged and exclusive breastfeeding on child height, weight, adiposity, and blood pressure at age 6.5 y: Evidence from a large randomized trial. The American Journal of Clinical Nutrition, 86(6), 1717-1721.

Kramer, M. S., & Kakuma, R. (2012). Optimal duration of exclusive breastfeeding (Review). The Cochrane Library.

279 Kratzer, L., & Hodgins, S. (1999). A typology of offenders: A test of Moffitt's theory among males and females from childhood to age 30. Criminal Behaviour and Mental Health, 9(1), 57-73.

Kuo, A. A., Inkelas, M., Slusser, W. M., Maidenberg, M., & Halfon, N. (2011). Introduction of solid food to young infants. Maternal and Child Health Journal, 15(8), 1185-1194.

Kwok, M. K., Leung, G. M., & Schooling, C. M. (2013). Breast feeding and early adolescent behaviour, self-esteem and depression: Hong Kong's ‘Children of 1997’ birth cohort. Archives of Disease in Childhood, 98(11), 887-894.

Ladomenou, F., Moschandreas, J., Kafatos, A., Tselentis, Y., & Galanakis, E. (2010). Protective effect of exclusive breastfeeding against infections during infancy: A prospective study. Archives of Disease in Childhood, 95(12), 1004-1008.

Lakatos, K., Toth, I., Nemoda, Z., Ney, K., Sasvari-Szekely, M., & Gervai, J. (2000). Dopamine D4 receptor (DRD4) gene polymorphism is associated with attachment disorganization in infants. Molecular Psychiatry, 5(6), 633-637.

Leerkes, E. M. (2011). Maternal sensitivity during distressing tasks: A unique predictor of attachment security. Infant Behavior & Development, 34(3), 443-446.

Lind, J. N., Li, R., Perrine, C. G., & Schieve, L. A. (2014). Breastfeeding and later psychosocial development of children at 6 years of age. Pediatrics, 134(Supplement 1), S36-S41.

Linnet, K. M., Wisborg, K., Agerbo, E., Secher, N. J., Thomsen, P. H., & Henriksen, T. B. (2006). Gestational age, birth weight, and the risk of hyperkinetic disorder. Archives of Disease in Childhood, 91(8), 655-660.

Liu, J. (2004). Prenatal and perinatal complications as predispositions to externalizing behavior. Journal of Prenatal & Perinatal Psychology & Health, 18(4), 301.

Liu, J. (2011). Early health risk factors for violence: Conceptualization, evidence, and implications. Aggression and Violent Behavior, 16(1), 63-73.

Liu, J., Hanlon, A., Ma, C., Zhao, S. R., Cao, S., & Compher, C. (2014). Low blood zinc, iron, and other sociodemographic factors associated with behavior problems in preschoolers. Nutrients, 6(2), 530-545.

Liu, J., Leung, P., & Yang, A. (2013). Breastfeeding and active bonding protects against children’s internalizing behavior problems. Nutrients, 6(1), 76-89.

Liu, J., & Raine, A. (2006). The effect of childhood malnutrition on externalizing behavior. Current Opinion in Pediatrics, 18(5), 565-570.

280 Liu, J., Raine, A., Venables, P. H., & Mednick, S. A. (2004). Malnutrition at age 3 years and externalizing behavior problems at ages 8, 11, and 17 years. American Journal of Psychiatry, 161(11), 2005-2013.

Liu, J., Raine, A., Wuerker, A., Venables, P. H., & Mednick, S. (2009). The association of birth complications and externalizing behavior in early adolescents: Direct and mediating effects. Journal of Research on Adolescence, 19(1), 93-111.

Liu, J., & Wuerker, A. (2005). Biosocial bases of aggressive and violent behavior—implications for nursing studies. International Journal of Nursing Studies, 42(2), 229-241.

Lönnerdal, B. (2013). Bioactive proteins in breast milk. Journal of Paediatrics and Child Health, 49(S1), 1-7.

LoParo, D., & Waldman, I. (2014). Twins’ rearing environment similarity and childhood externalizing disorders: A test of the equal environments assumption. Behavior Genetics, 44(6), 606-613.

Lu, J., Faber, A., & Dubé, L. (2013). An evidence-based approach to the nutritional quality of home meals: Exploring emotional reinforcement and attachment style as underlying mechanisms. In V. R. Preedy, L. Hunter, & V. B. Patel (Eds.), Diet quality: An evidence –based approach (vol. 1) (pp. 27-40). Springer New York.

Luu, T. M., Vohr, B. R., Allan, W., Schneider, K. C., & Ment, L. R. (2011). Evidence for catch- up in cognition and receptive vocabulary among adolescents born very preterm. Pediatrics, 128(2), 313-322.

Lucas, A., Morley, R., Cole, T. J., Lister, G., & Leeson-Payne, C. (1992). Breast milk and subsequent intelligence quotient in children born preterm. The Lancet, 339(8788), 261- 264.

Lyons-Ruth, K., Conneil, D. B., Zoll, D., & Stahl, J. (1987). Infants at social risk: Relations among infant maltreatment, maternal behavior, and infant attachment behavior. Developmental Psychology, 23(2), 223-232.

Malik, A. I., Zai, C. C., Abu, Z., Nowrouzi, B., & Beitchman, J. H. (2012). The role of oxytocin and oxytocin receptor gene variants in childhood onset aggression. Genes, Brain and Behavior, 11(5), 545-551. ‐ Martin, N. W., Benyamin, B., Hansell, N. K., Montgomery, G. W., Martin, N. G., Wright, M. J., & Bates, T. C. (2011). Cognitive function in adolescence: Testing for interactions between breast-feeding and< i> FADS2 polymorphisms. Journal of the American Academy of Child & Adolescent Psychiatry, 50(1), 55-62.

281 Martin, R. M., Gunnell, D., & Smith, G. D. (2005). Breastfeeding in infancy and blood pressure in later life: Systematic review and meta-analysis. American Journal of Epidemiology, 161(1), 15-26.

Mason, D. A., & Frick, P. J. (1994). The heritability of antisocial behavior: A meta-analysis of twin and adoption studies. Journal of Psychopathology and Behavioral Assessment, 16(4), 301-323.

Mathers, J. C., & Hesketh, J. E. (2007). The biological revolution: Understanding the impact of SNPs on diet-cancer interrelationships. The Journal of Nutrition, 137(1), 253S-258S.

Mattson, M. P. (2003). Gene–diet interactions in brain aging and neurodegenerative disorders. Annals of Internal Medicine, 139(5_Part_2), 441-444.

McLaughlin, K. A., Zeanah, C. H., Fox, N. A., & Nelson, C. A. (2012). Attachment security as a mechanism linking foster care placement to improved mental health outcomes in previously institutionalized children. Journal of Child Psychology and Psychiatry, 53(1), 46-55.

Melnick, S. M., & Hinshaw, S. P. (2000). Emotion regulation and parenting in AD/HD and comparison boys: Linkages with social behaviors and peer preference. Journal of Abnormal Child Psychology, 28(1), 73-86.

Menting, B., Van Lier, P. A., & Koot, H. M. (2011). Language skills, peer rejection, and the development of externalizing behavior from kindergarten to fourth grade. Journal of Child Psychology and Psychiatry, 52(1), 72-79.

Merjonen, P., Jokela, M., Pulkki-Råback, L., Hintsanen, M., Raitakari, O. T., Viikari, J., & Keltikangas-Järvinen, L. (2011). Breastfeeding and offspring hostility in adulthood. Psychotherapy and Psychosomatics, 80(6), 371-373.

Merrell, K.W. (2003). Preschool and Kindergarten Behavior Scales-2nd Edition. Austin, TX: Pro-Ed.

Mick, E., Biederman, J., Prince, J., Fischer, M. J., & Faraone, S. V. (2002). Impact of low birth weight on attention-deficit hyperactivity disorder. Journal of Developmental & Behavioral Pediatrics, 23(1), 16-22.

Miles, D. R., & Carey, G. (1997). Genetic and environmental architecture on human aggression. Journal of Personality and Social Psychology, 72(1), 207.

Mimouni-Bloch, A., Kachevanskaya, A., Mimouni, F. B., Shuper, A., Raveh, E., & Linder, N. (2013). Breastfeeding may protect from developing attention-deficit/hyperactivity disorder. Breastfeeding Medicine, 8(4), 363-367.

282 Miner, J. L., & Clarke-Stewart, K. A. (2008). Trajectories of externalizing behavior from age 2 to age 9: Relations with gender, temperament, ethnicity, parenting, and rater. Developmental Psychology, 44(3), 771.

Moffitt, T. E. (1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100(4), 674.

Moffitt, T. E. (2005). The new look of behavioral genetics in developmental psychopathology: Gene-environment interplay in antisocial behaviors. Psychological Bulletin, 131(4), 533.

Moffitt, T. E. (2006). A review of research on the taxonomy of life-course-persistent versus adolescence-limited antisocial behavior. In F. T. Cullen, J. P. Wright, & K. R. Blevins (Eds.), Taking stock: The status of criminological theory (pp. 277–311). New Brunswick, NJ: Transaction.

Moffitt, T. E., & Caspi, A. (2001). Childhood predictors differentiate life-course persistent and adolescence-limited antisocial pathways among males and females. Development and Psychopathology, 13(02), 355-375.

Moffitt, T. E., Caspi, A., Dickson, N., Silva, P., & Stanton, W. (1996). Childhood-onset versus adolescent-onset antisocial conduct problems in males: Natural history from ages 3 to 18 years. Development and Psychopathology, 8(02), 399-424.

Moffitt, T. E., Caspi, A., Harrington, H., & Milne, B. J. (2002). Males on the life-course- persistent and adolescence-limited antisocial pathways: Follow-up at age 26 years. Development and Psychopathology, 14(01), 179-207.

Molteni, R., Barnard, R. J., Ying, Z., Roberts, C. K., & Gómez-Pinilla, F. (2002). A high-fat, refined sugar diet reduces hippocampal brain-derived neurotrophic factor, neuronal plasticity, and learning. Neuroscience, 112(4), 803-814.

Morales, E., Bustamante, M., Gonzalez, J. R., Guxens, M., Torrent, M., Mendez, M., ... & Sunyer, J. (2011). Genetic variants of the FADS gene cluster and ELOVL gene family, colostrums LC-PUFA levels, breastfeeding, and child cognition. PLoS One, 6(2), e17181.

Mota, N. R., Bau, C. H., Banaschewski, T., Buitelaar, J. K., Ebstein, R. P., Franke, B., ... & Asherson, P. (2013). Association between DRD2/DRD4 interaction and conduct disorder: A potential developmental pathway to alcohol dependence. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics, 162(6), 546-549.

Mulvaney, M. K., & Mebert, C. J. (2007). Parental corporal punishment predicts behavior problems in early childhood. Journal of Family Psychology, 21(3), 389-397.

Nagin, D. S., & Land, K. C. (1993). Age, criminal careers, and population heterogeneity: Specification and estimation of a nonparametric, mixed Poisson model. Criminology, 31(3), 327-362.

283 Nagin, D., & Tremblay, R. E. (1999). Trajectories of boys' physical aggression, opposition, and hyperactivity on the path to physically violent and nonviolent juvenile delinquency. Child Development, 70(5), 1181-1196.

Nagin, D. S., & Tremblay, R. E. (2001). Parental and early childhood predictors of persistent physical aggression in boys from kindergarten to high school. Archives of General Psychiatry, 58(4), 389-394.

Nassar, M. F., Younis, N. T., El Arab, S. E., & Fawzi, F. A. (2011). Neuro developmental outcome and brain derived neurotrophic factor level in relation to feeding practice in early infancy. Maternal &‐ Child Nutrition, 7(2), 188-197. ‐ ‐ Nelson, E. E., & Panksepp, J. (1998). Brain substrates of infant–mother attachment: Contributions of opioids, oxytocin, and norepinephrine. Neuroscience & Biobehavioral Reviews, 22(3), 437-452.

Newton, E. R. (2004). Breastmilk: The gold standard. Clinical Obstetrics and Gynecology, 47(3), 632-642.

Niegel, S., Ystrom, E., Hagtvet, K. A., & Vollrath, M. E. (2008). Difficult temperament, breastfeeding, and their mutual prospective effects: The Norwegian Mother and Child Cohort Study. Journal of Developmental & Behavioral Pediatrics, 29(6), 458-462.

Nigg, J. T., & Breslau, N. (2007). Prenatal smoking exposure, low birth weight, and disruptive behavior disorders. Journal of the American Academy of Child & Adolescent Psychiatry, 46(3), 362-369.

Nissen, E., Uvnäs-Moberg, K., Svensson, K., Stock, S., Widström, A. M., & Winberg, J. (1996). Different patterns of oxytocin, prolactin but not cortisol release during breastfeeding in women delivered by caesarean section or by the vaginal route. Early Human Development, 45(1), 103-118.

Northstone, K., Joinson, C., Emmett, P., Ness, A., & Paus, T. (2011). Are dietary patterns in childhood associated with IQ at 8 years of age? A population-based cohort study. Journal of Epidemiology and Community Health, 66(7), 624-628.

Nyaradi, A., Li, J., Hickling, S., Foster, J., & Oddy, W. H. (2013). The role of nutrition in children's neurocognitive development, from pregnancy through childhood. Frontiers in Human Neuroscience, 7.

O'Connor, E. E., Collins, B. A., & Supplee, L. (2012). Behavior problems in late childhood: The roles of early maternal attachment and teacher–child relationship trajectories. Attachment & Human Development, 14(3), 265-288.

284 Oddy, W. H., Kendall, G. E., Li, J., Jacoby, P., Robinson, M., de Klerk, N. H., ... & Stanley, F. J. (2010). The long-term effects of breastfeeding on child and adolescent mental health: A pregnancy cohort study followed for 14 years. The Journal of Pediatrics, 156(4), 568- 574.

Oddy, W. H., Robinson, M., Ambrosini, G. L., de Klerk, N. H., Beilin, L. J., Silburn, S. R., ... & Stanley, F. J. (2009). The association between dietary patterns and mental health in early adolescence. Preventive Medicine, 49(1), 39-44.

Oddy, W. H., Robinson, M., Kendall, G. E., Li, J., Zubrick, S. R., & Stanley, F. J. (2011). Breastfeeding and early child development: A prospective cohort study. Acta Paediatrica, 100(7), 992-999.

Odgers, C. L., Moffitt, T. E., Broadbent, J. M., Dickson, N., Hancox, R. J., Harrington, H., ... & Caspi, A. (2008). Female and male antisocial trajectories: From childhood origins to adult outcomes. Development and Psychopathology, 20(2), 673.

Oh, S. Y., Ahn, H., Chang, N., Kang, M. H., & V Oh, J. (2013). Dietary patterns and weight status associated with behavioural problems in young children. Public Health Nutrition, 1-7.

Olds, D. L. (2006). The nurse–family partnership: An evidence based preventive intervention. Infant Mental Health Journal, 27(1), 5-25. ‐ Ordovas, J. M. (2006). Genetic interactions with diet influence the risk of cardiovascular disease. The American Journal of Clinical Nutrition, 83(2), 443S-446S.

Owen, M., & Cox, M. (1997). Marital conflict and the development of infant-parent attachment relationships. Journal of Family Psychology, 11(2), 152-164.

Panfile, T. M., & Laible, D. J. (2012). Attachment security and child's empathy: The mediating role of emotion regulation. Merrill-Palmer Quarterly, 58(1), 1-21.

Papp, L. M. (2013). Longitudinal associations between breastfeeding and observed mother–child interaction qualities in early childhood. Child: Care, Health and Development.

Park, S., Cho, S. C., Hong, Y. C., Oh, S. Y., Kim, J. W., Shin, M. S., ... & Bhang, S. Y. (2012). Association between dietary behaviors and attention-deficit/hyperactivity disorder and learning disabilities in school-aged children. Psychiatry Research, 198(3), 468-476.

Pearl, A. M., French, B. F., Dumas, J. E., Moreland, A. D., & Prinz, R. (2014). Bidirectional effects of parenting quality and child externalizing behavior in predominantly single parent, under-resourced African American families. Journal of Child and Family Studies, 23(2), 177-188.

285 Pearson, R. M., Lightman, S. L., & Evans, J. (2011). The impact of breastfeeding on mothers’ attentional sensitivity towards infant distress. Infant Behavior and Development, 34(1), 200-205.

Pederson, D. R., Moran, G., Sitko, C., Campbell, K., Ghesquire, K., & Acton, H. (1990). Maternal sensitivity and the security of infant mother attachment: A Q sort study. Child Development, 61(6), 1974-1983. ‐ ‐ Perrine, C. G., Galuska, D. A., Thompson, F. E., & Scanlon, K. S. (2014). Breastfeeding duration is associated with child diet at 6 years. Pediatrics, 134(Supplement 1), S50-S55.

Petkovsek, M. A., Boutwell, B. B., Beaver, K. M., & Barnes, J. C. (2014). Prenatal smoking and genetic risk: Examining the childhood origins of externalizing behavioral problems. Social Science & Medicine, 111, 17-24.

Plybon, L. E., & Kliewer, W. (2001). Neighborhood types and externalizing behavior in urban school-age children: Tests of direct, mediated, and moderated effects. Journal of Child and Family Studies, 10(4), 419-437.

Pingault, J. B., Côté, S. M., Galéra, C., Genolini, C., Falissard, B., Vitaro, F., & Tremblay, R. E. (2012). Childhood trajectories of inattention, hyperactivity and oppositional behaviors and prediction of substance abuse/dependence: A 15-year longitudinal population-based study. Molecular Psychiatry, 18(7), 806-812.

Piquero, A. (2001). Testing Moffitt's neuropsychological variation hypothesis for the prediction of life-course persistent offending. Psychology, Crime and Law, 7(1-4), 193-215.

Piquero, A. R. (2008). Taking stock of developmental trajectories of criminal activity over the life course. In The long view of crime: A synthesis of longitudinal research (pp. 23-78). Springer New York.

Piquero, A. R., Carriaga, M. L., Diamond, B., Kazemian, L., & Farrington, D. P. (2012). Stability in aggression revisited. Aggression and Violent Behavior, 17(4), 365-372.

Piquero, A. R., Farrington, D. P., Nagin, D. S., & Moffitt, T. E. (2010). Trajectories of offending and their relation to life failure in late middle age: Findings from the Cambridge Study in Delinquent Development. Journal of Research in Crime and Delinquency, 47(2), 151- 173.

Piquero, A. R., & White, N. A. (2003). On the relationship between cognitive abilities and life- course-persistent offending among a sample of African Americans: A longitudinal test of Moffitt's hypothesis. Journal of Criminal Justice, 31(5), 399-409.

Planalp, E. M., & Braungart-Rieker, J. M. (2013). Temperamental precursors of infant attachment with mothers and fathers. Infant Behavior and Development, 36(4), 796-808.

286 Polaschek, D. L. (2011). Many sizes fit all: A preliminary framework for conceptualizing the development and provision of cognitive–behavioral rehabilitation programs for offenders. Aggression and Violent Behavior, 16(1), 20-35.

Polderman, T. J., Posthuma, D., De Sonneville, L. M., Verhulst, F. C., & Boomsma, D. I. (2006). Genetic analyses of teacher ratings of problem behavior in 5-year-old twins. Twin Research and Human Genetics, 9(01), 122-130.

Propper, C., Willoughby, M., Halpern, C. T., Carbone, M. A., & Cox, M. (2007). Parenting quality, DRD4, and the prediction of externalizing and internalizing behaviors in early childhood. Developmental Psychobiology, 49(6), 619-632.

Qawasmi, A., Landeros-Weisenberger, A., Leckman, J. F., & Bloch, M. H. (2012). Meta- analysis of long-chain polyunsaturated fatty acid supplementation of formula and infant cognition. Pediatrics, 129(6), 1141-1149.

Quigley, M. A., Hockley, C., Carson, C., Kelly, Y., Renfrew, M. J., & Sacker, A. (2012). Breastfeeding is associated with improved child cognitive development: A population- based cohort study. The Journal of Pediatrics, 160(1), 25-32.

Quinn, P. J., O'Callaghan, M., Williams, G. M., Najman, J. M., Andersen, M. J., & Bor, W. (2001). The effect of breastfeeding on child development at 5 years: A cohort study. Journal of Paediatrics and Child Health, 37(5), 465-469.

Raby, K. L., Cicchetti, D., Carlson, E. A., Cutuli, J. J., Englund, M. M., & Egeland, B. (2012). Genetic and caregiving-based contributions to infant attachment unique associations with distress reactivity and attachment security. Psychological Science, 23(9), 1016-1023.

Raine, A. (2002). Annotation: The role of prefrontal deficits, low autonomic arousal, and early health factors in the development of antisocial and aggressive behavior in children. Journal of Child Psychology and Psychiatry, 43(4), 417-434.

Raine, A., Brennan, P., & Mednick, S. A. (1994). Birth complications combined with early maternal rejection at age 1 year predispose to violent crime at age 18 years. Archives of General Psychiatry, 51(12), 984-988.

Raine, A., Brennan, P., & Mednick, S. A. (1997). Interaction between birth complications and early maternal rejection in predisposing individuals to adult violence: Specificity to serious, early-onset violence. American Journal of Psychiatry, 154, 1265-1271.

Raine, A., Mellingen, K., Liu, J., & Venables, P., & Mednick, S.A. (2003). Effects of environmental enrichment at ages 3–5 years on schizotypal personality and antisocial behavior at ages 17 and 23 years. American Journal of Psychiatry, 160, 1627-1635.

287 Raine, A., Moffitt, T. E., Caspi, A., Loeber, R., Stouthamer-Loeber, M., & Lynam, D. (2005). Neurocognitive impairments in boys on the life-course persistent antisocial path. Journal of Abnormal Psychology, 114(1), 38.

Ramchandani, P. G., Domoney, J., Sethna, V., Psychogiou, L., Vlachos, H., & Murray, L. (2013). Do early father–infant interactions predict the onset of externalising behaviours in young children? Findings from a longitudinal cohort study. Journal of Child Psychology and Psychiatry, 54(1), 56-64.

Rapoport, J. L., & Gogtay, N. (2007). Brain neuroplasticity in healthy, hyperactive and psychotic children: Insights from neuroimaging. Neuropsychopharmacology, 33(1), 181-197.

Ratchford, M., & Beaver, K. M. (2009). Neuropsychological deficits, low self-control, and delinquent involvement: Toward a biosocial explanation of delinquency. Criminal Justice and Behavior, 36(2), 147-162.

Reif, A., Rösler, M., Freitag, C. M., Schneider, M., Eujen, A., Kissling, C., ... & Retz, W. (2007). Nature and nurture predispose to violent behavior: Serotonergic genes and adverse childhood environment. Neuropsychopharmacology, 32(11), 2375-2383.

Reynolds, A. (2001). Breastfeeding and brain development. Pediatric Clinics of North America, 48(1), 159-171.

Reynolds, D., Hennessy, E., & Polek, E. (2014). Is breastfeeding in infancy predictive of child mental well being and protective against obesity at 9 years of age?. Child: Care, Health and Development. ‐ Rhee, S. H., & Waldman, I. D. (2002). Genetic and environmental influences on antisocial behavior: A meta-analysis of twin and adoption studies. Psychological Bulletin, 128(3), 490.

Rhoades, B. L., Warren, H. K., Domitrovich, C. E., & Greenberg, M. T. (2011). Examining the link between preschool social–emotional competence and first grade academic achievement: The role of attention skills. Early Childhood Research Quarterly, 26(2), 182-191.

Rice, F., Harold, G. T., Boivin, J., Van den Bree, M., Hay, D. F., & Thapar, A. (2010). The links between prenatal stress and offspring development and psychopathology: Disentangling environmental and inherited influences. Psychological Medicine, 40(02), 335-345.

Riggs, N. R., Blair, C. B., & Greenberg, M. T. (2004). Concurrent and 2-year longitudinal relations between executive function and the behavior of 1st and 2nd grade children. Child Neuropsychology, 9(4), 267-276.

288 Riggs, N. R., Spruijt-Metz, D., Sakuma, K. L., Chou, C. P., & Pentz, M. A. (2010). Executive cognitive function and food intake in children. Journal of Nutrition Education and Behavior, 42(6), 398-403.

Riva Crugnola, C., Tambelli, R., Spinelli, M., Gazzotti, S., Caprin, C., & Albizzati, A. (2011). Attachment patterns and emotion regulation strategies in the second year. Infant Behavior and Development, 34(1), 136-151.

Rizzi, T. S., van der Sluis, S., Derom, C., Thiery, E., van Kesteren, R. E., Jacobs, N., ... & Posthuma, D. (2013). FADS2 genetic variance in combination with fatty acid intake might alter composition of the fatty acids in brain. PloS One, 8(6), e68000.

Rocque, M., Welsh, B. C., & Raine, A. (2012). Biosocial criminology and modern crime prevention. Journal of Criminal Justice, 40(4), 306-312.

Rodgers, J. L., Buster, M., & Rowe, D. C. (2001). Genetic and environmental influences on delinquency: DF analysis of NLSY kinship data. Journal of Quantitative Criminology, 17(2), 145-168.

Rodgers, J. L., & Kohler, H. P. (2005). Reformulating and simplifying the DF analysis model. Behavior Genetics, 35(2), 211-217.

Rodgers, J. L., Rowe, D. C., & Li, C. (1994). Beyond nature versus nurture: DF analysis of nonshared influences on problem behaviors. Developmental Psychology, 30(3), 374.

Roisman, G. I., & Fraley, R. C. (2008). A behavior-genetic study of parenting quality, infant attachment security, and their covariation in a nationally representative sample. Developmental Psychology, 44(3), 831.

Rowe, M. L., Raudenbush, S. W., & Goldin Meadow, S. (2012). The pace of vocabulary growth helps predict later vocabulary skill. Child Development, 83(2), 508-525. ‐ Rubin, K. H., Burgess, K. B., Dwyer, K. M., & Hastings, P. D. (2003). Predicting preschoolers' externalizing behaviors from toddler temperament, conflict, and maternal negativity. Developmental Psychology, 39(1), 164.

Ryan, A. S., Astwood, J. D., Gautier, S., Kuratko, C. N., Nelson, E. B., & Salem Jr, N. (2010). Effects of long-chain polyunsaturated fatty acid supplementation on neurodevelopment in childhood: A review of human studies. Prostaglandins, Leukotrienes and Essential Fatty Acids (PLEFA), 82(4), 305-314.

Saarinen, S., Fontell, T., Vuontela, V., Carlson, S., & Aronen, E. T. (2014). Visuospatial working memory in 7-to 12-year-old children with disruptive behavior disorders. Child Psychiatry & Human Development, 1-10.

289 Sabuncuoglu, O., Orengul, C., Bikmazer, A., & Yilmaz Kaynar, S. (2014). Breastfeeding and parafunctional oral habits in children with and without Attention-Deficit/Hyperactivity Disorder. Breastfeeding Medicine, 9(5), 244-250.

Sampson, R. J., & Laub, J. H. (1993). Crime in the making: Pathways and turning points through life. Cambridge, MA: Harvard University Press.

Saudino, K. J., Ronald, A., & Plomin, R. (2005). The etiology of behavior problems in 7-year- old twins: Substantial genetic influence and negligible shared environmental influence for parent ratings and ratings by same and different teachers. Journal of Abnormal Child Psychology, 33(1), 113-130.

Schaeffer, C. M., Petras, H., Ialongo, N., Poduska, J., & Kellam, S. (2003). Modeling growth in boys' aggressive behavior across elementary school: Links to later criminal involvement, conduct disorder, and antisocial personality disorder. Developmental Psychology, 39(6), 1020.

Scheres, A., Oosterlaan, J., Geurts, H., Morein-Zamir, S., Meiran, N., Schut, H., ... & Sergeant, J. A. (2004). Executive functioning in boys with ADHD: Primarily an inhibition deficit?. Archives of Clinical Neuropsychology, 19(4), 569-594.

Schmied, V., & Lupton, D. (2001). Blurring the boundaries: Breastfeeding and maternal subjectivity. Sociology of Health & Illness, 23(2), 234-250.

Schoemaker, K., Mulder, H., Deković, M., & Matthys, W. (2013). Executive functions in preschool children with externalizing behavior problems: A meta-analysis. Journal of Abnormal Child Psychology, 41(3), 457-471.

Schoenthaler, S. J. (1983). Diet and crime: An empirical examination of the value of nutrition in the control and treatment of incarcerated juvenile offenders. International Journal of Biosocial Research, 4(1), 25-39.

Schoenthaler, S. J. (1985). Diet and delinquency: Empirical testing of seven theories. International Journal of Biosocial Research, 7(2), 108-131.

Schoenthaler, S., Amos, S., Doraz, W., Kelly, M. A., Muedeking, G., & Wakefield, J. (1997). The effect of randomized vitamin-mineral supplementation on violent and non-violent antisocial behavior among incarcerated juveniles. Journal of Nutritional & Environmental Medicine, 7(4), 343-352.

Schoenthaler, S. J., & Bier, I. D. (2000). The effect of vitamin-mineral supplementation on juvenile delinquency among american schoolchildren: A randomized, double-blind placebo-controlled trial. The Journal of Alternative and Complementary Medicine, 6(1), 7-17.

290 Schuchardt, J. P., Huss, M., Stauss-Grabo, M., & Hahn, A. (2010). Significance of long-chain polyunsaturated fatty acids (PUFAs) for the development and behaviour of children. European Journal of Pediatrics, 169(2), 149-164.

Schulze, P. A., & Carlisle, S. A. (2010). What research does and doesn’t say about breastfeeding: A critical review. Early Child Development and Care, 180(6), 703-718.

Scott, J. A., Chih, T. Y., & Oddy, W. H. (2012). Food variety at 2 years of age is related to duration of breastfeeding. Nutrients, 4(10), 1464-1474.

Shapiro, D. L., & Levendosky, A. A. (1999). Adolescent survivors of childhood sexual abuse: The mediating role of attachment style and coping in psychological and interpersonal functioning. Child Abuse & Neglect, 23(11), 1175-1191.

Shaw, D. S., Keenan, K., & Vondra, J. I. (1994). Developmental precursors of externalizing behavior: ages 1 to 3. Developmental Psychology, 30(3), 355.

Shaw, D. S., Lacourse, E., & Nagin, D. S. (2005). Developmental trajectories of conduct problems and hyperactivity from ages 2 to 10. Journal of Child Psychology and Psychiatry, 46(9), 931-942.

Shelton, K. H., Collishaw, S., Rice, F. J., Harold, G. T., & Thapar, A. (2011). Using a genetically informative design to examine the relationship between breastfeeding and childhood conduct problems. European Child & Adolescent Psychiatry, 20(11-12), 571-579.

Shenassa, E. D., Paradis, A. D., Dolan, S. L., Wilhelm, C. S., & Buka, S. L. (2012). Childhood impulsive behavior and problem gambling by adulthood: A 30 year prospective community based study. Addiction, 107(1), 160-168. ‐ Silk, T. J., Vance, A.,‐ Rinehart, N., Bradshaw, J. L., & Cunnington, R. (2009). White matter abnormalities in attention deficit hyperactivity disorder: A diffusion tensor imaging study. Human Brain Mapping, 30(9), 2757-2765. ‐

Simmons, R. J., Goldberg, S., Washington, J., Fischer-Fay, A., & Maclusky, I. (1995). Infant- mother attachment and nutrition in children with cystic fibrosis. Journal of Developmental & Behavioral Pediatrics, 16(3), 183-186.

Singh, M. (2005). Essential fatty acids, DHA and human brain. The Indian Journal of Pediatrics, 72(3), 239-242.

Sinn, N. (2008). Nutritional and dietary influences on attention deficit hyperactivity disorder. Nutrition Reviews, 66(10), 558-568.

Sinn, N., & Bryan, J. (2007). Effect of supplementation with polyunsaturated fatty acids and micronutrients on learning and behavior problems associated with child ADHD. Journal of Developmental & Behavioral Pediatrics, 28(2), 82-91.

291 Smith, M. M., Durkin, M., Hinton, V. J., Bellinger, D., & Kuhn, L. (2003). Initiation of breastfeeding among mothers of very low birth weight infants. Pediatrics, 111(6), 1337- 1342.

Stam, J., Sauer, P. J., & Boehm, G. (2013). Can we define an infant's need from the composition of human milk?. The American Journal of Clinical Nutrition, 98(2), 521S-528S.

Stattin, H., & Magnusson, D. (1989). The role of early aggressive behavior in the frequency, seriousness, and types of later crime. Journal of Consulting and Clinical Psychology, 57(6), 710.

Steer, C. D., Smith, G. D., Emmett, P. M., Hibbeln, J. R., & Golding, J. (2010). FADS2 polymorphisms modify the effect of breastfeeding on child IQ. PLoS One, 5(7), e11570.

Storebo, O. J., Rasmussen, P. D., & Simonsen, E. (2013). Association between insecure attachment and ADHD: Environmental mediating factors. Journal of Attention Disorders, 1-10.

Strathearn, L., Fonagy, P., Amico, J., & Montague, P. R. (2009). Adult attachment predicts maternal brain and oxytocin response to infant cues. Neuropsychopharmacology, 34(13), 2655-2666.

Strathearn, L., Iyengar, U., Fonagy, P., & Kim, S. (2012). Maternal oxytocin response during mother–infant interaction: Associations with adult temperament. Hormones and Behavior, 61(3), 429-435.

Stuebe, A. M., Grewen, K., & Meltzer-Brody, S. (2013). Association between maternal mood and oxytocin response to breastfeeding. Journal of Women's Health, 22(4), 352-361.

Sullivan, S. A., & Birch, L. L. (1994). Infant dietary experience and acceptance of solid foods. Pediatrics, 93(2), 271-277.

Susman-Stillman, A., Kalkoske, M., Egeland, B., & Waldman, I. (1996). Infant temperament and maternal sensitivity as predictors of attachment security. Infant Behavior and Development, 19(1), 33-47.

Taki, Y., Hashizume, H., Sassa, Y., Takeuchi, H., Asano, M., Asano, K., & Kawashima, R. (2010). Breakfast staple types affect brain gray matter volume and cognitive function in healthy children. PLoS One, 5(12), e15213.

Tanaka, K., Kon, N., Ohkawa, N., Yoshikawa, N., & Shimizu, T. (2009). Does breastfeeding in the neonatal period influence the cognitive function of very-low-birth-weight infants at 5 years of age?. Brain and Development, 31(4), 288-293.

Teisl, M., & Cicchetti, D. (2008). Physical abuse, cognitive and emotional processes, and aggressive/disruptive behavior problems. Social Development, 17(1), 1-23.

292 Tharner, A., Luijk, M. P., Raat, H., IJzendoorn, M. H., Bakermans-Kranenburg, M. J., Moll, H. A., ... & Tiemeier, H. (2012). Breastfeeding and its relation to maternal sensitivity and infant attachment. Journal of Developmental & Behavioral Pediatrics, 33(5), 396-404.

Thapar, A., Rice, F., Hay, D., Boivin, J., Langley, K., van den Bree, M., ... & Harold, G. (2009). Prenatal smoking might not cause attention-deficit/hyperactivity disorder: Evidence from a novel design. Biological Psychiatry, 66(8), 722-727.

Theodore, R. F., Thompson, J., Waldie, K. E., Wall, C., Becroft, D. M., Robinson, E., ... & Mitchell, E. A. (2009). Dietary patterns and intelligence in early and middle childhood. Intelligence, 37(5), 506-513.

Thomas, C. R., & Pope, K. (Eds.). (2012). The origins of antisocial behavior: A developmental perspective. Oxford University Press.

Thompson, R., Tabone, J. K., Litrownik, A. J., Briggs, E. C., Hussey, J. M., English, D. J., & Dubowitz, H. (2010). Early adolescent risk behavior outcomes of childhood externalizing behavioral trajectories. Journal of Early Adolescence, 31(2), 234-257.

Tibbetts, S. G., & Piquero, A. R. (1999). The influence of gender, low birth weight, and disadvantaged environment in predicting early onset of offending: A test of Moffitt's interactional hypothesis. Criminology, 37(4), 843-878.

Timmermans, M., Van Lier, P. A., & Koot, H. M. (2008). Which forms of child/adolescent externalizing behaviors account for late adolescent risky sexual behavior and substance use?. Journal of Child Psychology and Psychiatry, 49(4), 386-394.

Torres, N., Maia, J., Veríssimo, M., Fernandes, M., & Silva, F. (2012). Attachment security representations in institutionalized children and children living with their families: Links to problem behaviour. Clinical Psychology & Psychotherapy, 19(1), 25-36.

Tremblay, R. E. (2015). Antisocial behavior before the age–crime curve: Can developmental criminology continue to ignore developmental origins?. In J. Morizot & L. Kazemian (Eds.), The development of criminal and antisocial behavior (pp. 39-49). Springer International Publishing.

Tremblay, R. E., Nagin, D. S., Séguin, J. R., Zoccolillo, M., Zelazo, P. D., Boivin, M., ... & Japel, C. (2004). Physical aggression during early childhood: Trajectories and predictors. Pediatrics, 114(1), e43-e50.

Tremblay, R. E., Pihl, R. O., Vitaro, F., & Dobkin, P. L. (1994). Predicting early onset of male antisocial behavior from preschool behavior. Archives of General Psychiatry, 51(9), 732- 739.

293 Trombini, E., Baldaro, B., Bertaccini, R., Mattei, C., Montebarocci, O., & Rossi, N. (2003). Maternal attitudes and attachment styles in mothers of obese children. Perceptual and Motor Skills, 97(2), 613-620.

Turner, M. G., Hartman, J. L., & Bishop, D. M. (2007). The effects of prenatal problems, family functioning, and neighborhood disadvantage in predicting life-course-persistent offending. Criminal Justice and Behavior, 34(10), 1241-1261.

Van den Bergh, B. R., & Marcoen, A. (2004). High antenatal maternal anxiety is related to ADHD symptoms, externalizing problems, and anxiety in 8 and 9 year olds. Child Development, 75(4), 1085-1097. ‐ ‐ ‐ van der Geest, V., Blokland, A., & Bijleveld, C. (2009). Delinquent development in a sample of high-risk youth shape, content, and predictors of delinquent trajectories from age 12 to 32. Journal of Research in Crime and Delinquency, 46(2), 111-143. van der Mark, I. L., IJzendoorn, M. H. V., & Bakermans Kranenburg, M. J. (2002). Development of empathy in girls during the second year of life: Associations with parenting, attachment, and temperament. Social Development‐ , 11(4), 451-468.

Van Ijzendoorn, M. H., Vereijken, C. M., Bakermans Kranenburg, M. J., & Marianne Riksen Walraven, J. (2004). Assessing attachment security with the attachment Q sort: Meta analytic evidence for the validity of the observer‐ AQS.Child development, 75(4), 1188-‐ 1213. ‐ van Lier, P., Boivin, M., Dionne, G., Vitaro, F., Brendgen, M., Koot, H., ... & Perusse, D. (2007). Kindergarten children's genetic vulnerabilities interact with friends' aggression to promote children's own aggression. Journal of the American Academy of Child & Adolescent Psychiatry, 46(8), 1080-1087.

Vaughn, M. G., DeLisi, M., Beaver, K. M., & Wright, J. P. (2009). Identifying latent classes of behavioral risk based on early childhood: Manifestations of self-control. Youth Violence and Juvenile Justice, 7(1), 16-31.

Vaughn, M. G., DeLisi, M., Gunter, T., Fu, Q., Beaver, K. M., Perron, B. E., & Howard, M. O. (2011). The severe 5%: A latent class analysis of the externalizing behavior spectrum in the United States. Journal of Criminal Justice, 39(1), 75-80.

Vohr, B. R., Poindexter, B. B., Dusick, A. M., McKinley, L. T., Higgins, R. D., Langer, J. C., & Poole, W. K. (2007). Persistent beneficial effects of breast milk ingested in the neonatal intensive care unit on outcomes of extremely low birth weight infants at 30 months of age. Pediatrics, 120(4), e953-e959.

294 Wade, S. L., Cassedy, A., Walz, N. C., Taylor, H. G., Stancin, T., & Yeates, K. O. (2011). The relationship of parental warm responsiveness and negativity to emerging behavior problems following traumatic brain injury in young children. Developmental Psychology, 47(1), 119.

Wadsby, M., Nelson, N., Ingemansson, F., Samuelsson, S., & Leijon, I. (2014). Behaviour problems and cortisol levels in very-low-birth-weight children. Nordic Journal of Psychiatry, 68(8), 626-632.

Waters, E., & Deane, K. E. (1985). Defining and assessing individual differences in attachment relationships: Q-methodology and the organization of behavior in infancy and early childhood. Monographs of the Society for Research in Child Development, 50, 41-65.

West, K. K., Mathews, B. L., & Kerns, K. A. (2013). Mother–child attachment and cognitive performance in middle childhood: An examination of mediating mechanisms. Early Childhood Research Quarterly, 28(2), 259-270.

Whalley, L. J., Deary, I. J., Starr, J. M., Wahle, K. W., Rance, K. A., Bourne, V. J., & Fox, H. C. (2008). n–3 Fatty acid erythrocyte membrane content, APOE ε4, and cognitive variation: An observational follow-up study in late adulthood. The American Journal of Clinical Nutrition, 87(2), 449-454.

Wigg, N. R., Tong, S., McMichael, A. J., Baghurst, P. A., Vimpani, G., & Roberts, R. (1998). Does breastfeeding at six months predict cognitive development?. Australian and New Zealand Journal of Public Health, 22(2), 232-236.

Wiles, N. J., Northstone, K., Emmett, P., & Lewis, G. (2007). ‘Junk food’diet and childhood behavioural problems: Results from the ALSPAC cohort. European Journal of Clinical Nutrition, 63(4), 491-498.

Wolff, M. S., & Ijzendoorn, M. H. (1997). Sensitivity and attachment: A meta analysis on parental antecedents of infant attachment. Child Development, 68(4), 571-591. ‐ Woo, H. D., Kim, D. W., Hong, Y. S., Kim, Y. M., Seo, J. H., Choe, B. M., ... & Kim, J. (2014). Dietary patterns in children with Attention Deficit/Hyperactivity Disorder (ADHD). Nutrients, 6(4), 1539-1553.

Wright, J., Beaver, K., DeLisi, M., & Vaughn, M. (2008). Evidence of negligible parenting influences on self control, delinquent peers, and delinquency in a sample of twins. Justice Quarterly, 25(3), 544-569. ‐ Wright, K., Coverston, C., Tiedeman, M., & Abegglen, J. A. (2006). Formula supplemented with docosahexaenoic acid (DHA) and arachidonic acid (ARA): A critical review of the research. Journal for Specialists in Pediatric Nursing, 11(2), 100-112.

295 Wright, J.P., Tibbetts, S.G., & Daigle, L. E. (2008). Criminals in the making: Criminality across the life course. Sage Publications, Inc.

Yorifuji, T., Kubo, T., Yamakawa, M., Kato, T., Inoue, S., Tokinobu, A., & Doi, H. (2014). Breastfeeding and behavioral development: A nationwide longitudinal survey in Japan. The Journal of Pediatrics, 164(5), 1019-1025.

Zaalberg, A., Nijman, H., Bulten, E., Stroosma, L., & van der Staak, C. (2010). Effects of nutritional supplements on aggression, rule breaking, and psychopathology among young adult prisoners. Aggressive Behavior, 36(2), 117-126. ‐ Zai, C. C., Muir, K. E., Nowrouzi, B., Shaikh, S. A., Choi, E., Berall, L., ... & Kennedy, J. L. (2012). Possible genetic association between vasopressin receptor 1B and child aggression. Psychiatry Research, 200(2), 784-788.

296 BIOGRAPHICAL SKETCH

Dylan B. Jackson

Dylan Jackson received his Bachelor’s degree in psychology from Brigham Young University, and his Master’s degree in criminology and criminal justice from Florida State University. His research interests include developmental/biosocial criminology, criminological theory, child development, nutrition/health, and neuropsychology. His research has appeared in a diverse set of journals, including Journal of Research in Crime and Delinquency, Psychiatry Research,

Journal of Criminal Justice, International Journal of Environmental Research and Public

Health, Youth Violence and Juvenile Justice, and Psychiatric Quarterly. In the fall of 2015, he will join the faculty in the Department of Criminal Justice at the University of Texas at San

Antonio.

297