UNDERSTANDING AND QUANTIFYING THE ROLES OF PERCEIVED SOCIAL

SUPPORT, PET ATTACHMENT, AND ADULT ATTACHMENT

IN ADULT PET OWNERS’ SENSE OF WELL-BEING

By

STEPHANIE CECILE LANGSTON

A dissertation submitted in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

WASHINGTON STATE UNIVERSITY Department of Educational Leadership, Sport Studies, & Educational/Counseling

AUGUST 2014

© Copyright by STEPHANIE LANGSTON, 2014 All Rights Reserved

© Copyright by STEPHANIE LANGSTON, 2014 All Rights Reserved

To the Faculty of Washington State University

The members of the Committee appointed to examine the dissertation of STEPHANIE LANGSTON find it satisfactory and recommend that it be accepted.

______Phyllis Erdman, Ph.D., Chair

______Laurie "Lali" McCubbin, Ph.D.

______Sarah Ullrich-French, Ph.D.

______Kathleen “Kathy" Ruby, Ph.D.

ii ACKNOWLEDGMENTS

My progress and accomplishments throughout my graduate career would not have been

possible without the commitment and support of several noteworthy individuals. I would like to

thank my advisors and mentors, Dr. Phyllis Erdman and Dr. Laurie “Lali” McCubbin, for

guiding and supporting me through this process, both personally and professionally. I also want

to extend my gratitude and respect to Dr. Sarah Ullrich-French and Dr. Kathy Ruby for their invaluable knowledge and enthusiasm for this project.

I have been graced by the , support, and power that come with being part of a

phenomenal unit. I want to thank my and for instilling in me a for

learning and education, and for their willingness to nurture a menagerie of pets throughout childhood. I am also grateful to my sister and her partner. In their love, as well as my , I learned how enriching a secure relationship can be, and I understood in a tangible way just how unstoppable a woman can be when she is encouraged to pursue her dreams and goals.

I would also like to thank my supervisors and colleagues at Palouse River Counseling,

Washington State’s Counseling and Testing Services, and Appalachian State University’s

Counseling Center. These individuals have had a profound impact on my personal and

professional growth as helper, and I have treasured becoming a member of these communities.

Finally, this journey would not have been nearly as fabulous without the companions,

both human and furry, I have had along the way. No-filter Fridays, dog walks, and constant support made this journey far more fun than I ever thought possible. I appreciate Conan, Pepper, and Pearl, my furry editors and lap warmers throughout the writing and research process. My cup runneth over and my heart is full. I am thankful for the roots that have kept me grounded, the passion in my mind, the fire in my soul, and the wings I can use to discover new things.

iii UNDERSTANDING AND QUANTIFYING THE ROLES OF PERCEIVED SOCIAL

SUPPORT, PET ATTACHMENT, AND ADULT ATTACHMENT

IN ADULT PET OWNERS’ SENSE OF WELL-BEING

Abstract

By Stephanie Langston, Ph. D. Washington State University August 2014

Chair: Phyllis Erdman, Ph.D.

Researchers have extolled the benefits of pet ownership. Most of the studies which have been published lack a theoretical lens through which to conceptualize the human-animal bond.

Attachment theory offers a framework for understanding the connection between humans and pets. This study applied to the human-animal bond in an effort to understand whether pet attachment (PAtt) is related to adult attachment (AAtt), subjective well-being

(SWB), and perceived social support (PSS). Beyond correlational analyses, this study also examined the degree to which individual differences in pet attachment can uniquely predict subjective well-being in adults who own pets.

Participants were 561 pet-owning adults who completed a demographic questionnaire, the Experiences in Close Relationships Scale (ECR), the Pet Attachment Questionnaire (PAQ), the Perceived Social Support from Friends and Family Scales (PSS), the Satisfaction With Life

Scale (SWLS), and the Positive and Negative Affect Schedule Expanded Form (PANAS-X).

Results indicated that Adult Attachment (AAtt) scales ( and avoidance) were inversely correlated with positive affect (PA) and satisfaction with life (SWLS), and were positively correlated with negative affect (NA). Perceived Social Support (PSS) was positively correlated with PA and SWLS and negatively correlated with NA. PSS was a significant

iv predictor of SWB, AAtt was a significant predictor of SWB, and PSS predicted AAtt. AAtt partially mediated the relationship between PSS and PA, fully mediated the relationship between

PSS and NA, and partially mediated the relationship between PSS and SWLS. No significant moderation effects were found for the interaction between AAtt and PSS in predicting participants’ SWB. Pet Attachment (PAtt) anxiety had significant correlations with AAtt anxiety and avoidance, while PAtt avoidance had a significant correlation with AAtt avoidance and no relationship with AAtt anxiety. PAtt avoidance was significantly associated with NA, but not

PSS, SWLS, or PA; PAtt anxiety was significantly correlated with PSS, SWLS, PA, and NA.

PAtt avoidance significantly predicted SWLS and PAtt anxiety significantly predicted NA. PSS was not found to be a significant predictor in either facet of PAtt, and PSS did not predict PAtt.

Interpretation and limitations of the findings, applied implications, and future directions are discussed.

v

TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS ……………………………………………………………….. iii

ABSTRACT ……………………………………………………………………………….. iv

LIST OF TABLES …………………………………………………………………………. ix

LIST OF FIGURES………………………………………………………………………… xi

CHAPTER

1. INTRODUCTION …………………………………………………………………. 1 Introduction ……………………………………………………………… 1 Review of the Literature……………………………………………………. 1 Gaps in the Research ……………………………………………………….. 3 The Current Study ………………………………………………………….. 4 Significance………………………………………………………………… 5 Purpose and Research Questions…………………………………………… 5 Hypotheses …………………………………………………………………. 6 2. REVIEW OF THE LITERATURE………………………………………………… 8 Attachment Theory…………………………………………………………. 8 Tenets of Attachment Theory ……………………………………… 8 Measuring and Quantifying Attachment …………………………… 13 Neural Pathways of Attachment………………………………...... 18 In Children …………………………………………………. 18 In Adults and Across the Lifespan …………………………. 22

vi Animal Models and Bonding ………………………………. 29 Benefits of the Human-Animal Bond (HAB) ……………………………… 30 Important Terminology in the Human-Animal Bond ……………… 32 Evidence of the Physiological Benefits of the HAB ……………….. 34 Effects on the Companion Animal …………………………. 40 Effects on Humans- Clinical Samples……………………… 41 Effects on Humans- Non-clinical Samples…………………. 42 Evidence of the Psychosocial Benefits of the HAB………………… 45 Effects on Humans- Clinical Samples……………………… 48 Effects on Humans- Non-clinical Samples ………………… 54 A Crossroads: The HAB and Attachment Theory………………………….. 63 Construct Equivalence?…………………………………………….. 63 Pets as Attachment Figures ………………………………………… 65 Attachment Behaviors ……………………………………… 66 Attachment Bonds ………………………………………….. 67 Grief and Loss ……………………………………………… 69 Operationalizing Pet Attachment…………………………………………… 71 Ainsworth’s Procedure ………………………….. 71 Constructing Self-Report Measures………………………………… 74 Current Trends in Research………………………………………… 84 Summary…….……………………………………………………………… 88 Gaps in Research…………………………………………………………… 90 Overview of the Present Study……………………………………………… 93

3. RESEARCH DESIGN AND METHODOLOGY………………………………….. 99 Introduction ………………………………………………………………… 99 Instruments …………………………………………………………………. 99

vii Participants …………………………………………………………………. 103 Procedure…………………………………………………………………… 107 Data Analyses………………………………………………………………. 108

4. RESULTS ………………………………………………………………………….. 114 Descriptive Statistics ……………………………………………………….. 114 Pearson Correlations ……………………………………………………….. 115 Tests of Hypotheses………………………………………………………… 117

5. DISCUSSION………………………………………………………………………. 156 Summary and Interpretation of Results…………………………………….. 156 Applied Implications ……………………………………………………….. 171 Limitations and Future Directions………………………………………….. 174 Conclusion………………………………………………………………….. 177

REFERENCES……………………………………………………………………………… 179

APPENDICES……………………………………………………………………………… 190

A. INFORMED CONSENT…………………………………………………………… 190

B. EMAIL RECRUITMENT FLYER ………………………………………………… 193

C. DEMOGRAPHIC QUESTIONNAIRE ……………………………………………. 194

viii LIST OF TABLES Page

1. Descriptive Statistics for Primary Constructs ……………………………………… 115

2. Correlations Between Primary Constructs …………………………………………. 116

3. Hierarchical Multiple Regression with PSS and Subjective Well-being ……………………………………………………………… 119

4. Model Fit for Hierarchical Multiple Regression with PSS and Subjective Well-being…………………………………………………….. 119

5. Hierarchical Multiple Regression with Adult Attachment and Subjective Well-being…………………………………………….. 122

6. Model Fit for Hierarchical Multiple Regression with AAtt and SWB………………………………………………………………… 122

7. Hierarchical Multiple Regression with PSS and Adult Attachment…………………………………………………………………… 124

8. Model Fit for Hierarchical Multiple Regression with PSS and AAtt …………………………………………………………………. 125

9. Hierarchical Multiple Regression with Perceived Social Support, Adult Attachment, and Subjective Well-being……………………. 128

10. Model Fit for Hierarchical Multiple Regression with PSS, AAtt, and SWB………………………………………………………….. 128

11. Hypothesis 6, Mediational Analysis, Baron and Kenny Steps (Kenny, 2013)………………………………………………………… 131

12. Hierarchical Multiple Regression with Perceived Social Support, Adult Attachment, and Subjective Well-being for Moderation, Hypothesis 7………………………………………….. 134

13. Model for Hierarchical Multiple Regression with PSS, AAtt, and SWB, for Moderation, Hypothesis 7………………………………. 135

14. Hierarchical Multiple Regression with Pet Attachment and Subjective Well-being …………………………………………………………. 140

15. Model Fit for Hierarchical Multiple Regression with PAtt and SWB ………………………………………………………………… 140

ix 16. Hierarchical Multiple Regression with PSS and Pet Attachment ………………………………………………………………… 144

17. Model for Hierarchical Multiple Regression with PSS and PAtt ………………………………………………………………….. 145

18. Hierarchical Multiple Regression with Perceived Social Support, Pet Attachment, and Subjective Well-being for Moderation, Hypothesis 13………………………………………… 151

19. Model for Hierarchical Multiple Regression with PSS, PAtt, and SWB for Moderation, Hypothesis 13 ……………………………… 152

x LIST OF FIGURES Page

1. Components of the Human Attachment Behavior System…………………………. 10

2. Additional Concepts Within Attachment Theory…………………………………… 12

3. Forms of Human-Animal Interaction………………………………………………. 33

4. Classifications of Animals………………………………………………………….. 33

5. Proposed Mechanisms of Action for Physiological Effects of HAB ………………. 40

6. Proposed Mechanisms for Psychosocial Benefits of HAB ………………………… 48

7. Summary of HAB Benefits ………………………………………………………… 63

8. Comparing Types of Bonds………………………………………………………… 71

9. Hypothesized Mediation Model of Perceived Social Support, Adult Attachment, and Subjective Well-being- Hypothesis 6……………………… 96

10. Hypothesized Moderation Model of Perceived Social Support, Adult Attachment, and Subjective Well-being- Hypothesis 7……………………… 97

11. Hypothesized Mediation Model of Perceived Social Support, Pet Attachment, and Subjective Well-being- Hypothesis 12……………………….. 97

12. Hypothesized Moderation Model of Perceived Social Support, Pet Attachment, and Subjective Well-being- Hypothesis 13……………………….. 98

13. Hypothesis 6, Mediation Diagram with Unstandardized Coefficients……………… 132

14. Moderation Model of Perceived Social Support, Adult Attachment, and Subjective Well-being- Hypothesis 7………………………………………………. 137

15. Pet Attachment Mediation Diagram with Unstandardized Coefficients …………… 150

16. Moderation Model of Perceived Social Support, Pet Attachment, and Subjective Well-being- Hypothesis 13 …………………………………………….. 155

xi

Dedication

This dissertation is dedicated to my family members, human and non-human, who taught me the meaning and value of unconditional support, and to Mitochondrial Eve, who instilled in us all the

power to love and connect with other beings.

xii CHAPTER ONE

“The therapist who conducts therapy with the pet seems less threatening, and the patient may reveal more of himself. Just as the therapist seems less forbidding and more human, the patient with the pet is perceived by others as more human and more treatable. These expectations create a positive cycle and the patient is treated in ways that encourage recovery.” (Beck et al., 1986, p. 63) Introduction

Animals and humans have coexisted for thousands of years. Evidence of a deep and enduring emotional bond has been present for nearly as long. A human skeleton holding a puppy was found in northern Israel 12,000 years old (Morrison, 2007). Animals have served humans as sources of food, tools for labor, companions, and more recently, as service assistants, which have all contributed to the development and investigation of the human-animal bond

(HAB). Today, more Americans own pets than have children (Flynn, 2000). Theories as to the causes, processes, and mechanisms of these relationships abound, and most fall within two realms: the biological and physiological underpinnings of the connection (Brickel, 1982; Shiloh,

Sorek, & Terkel, 2003), and the psychological and psychosocial bases of the bond between human beings and companion animals (Cline, 2010; Insel & Young, 2001; Jenkins, 1986; Walsh,

2009).

Review of the Literature

Scholars have found physiological evidence that the HAB is beneficial by reducing blood pressure and cortisol levels, increasing oxytocin levels, and activating a relaxation response (e.g.,

Baun et al., 1984; Cole et al., 2007; Lust et al., 2007; Miller et al., 2009). Multiple mechanisms have been proposed for these effects, including direct effects related to tactile contact with another living organism, classical conditioning theory, and competing response theory (Brickel,

1982; Shiloh, Sorek, & Terkel, 2003). In investigations of the psychosocial benefits of the HAB, researchers have found that individuals who interact with companion animals report higher self-

1 esteem, increased frequency and quality of interpersonal interactions, and increased pro-social behaviors (Antonacopoulos & Pychyl, 2010; Knight & Edwards, 2008; Yorke, Adams, & Coady,

2008). Authors often assumed these effects were related to social facilitation, psychodynamic roles of animals in relationships, or anthropomorphism (Cline, 2010; Insel & Young, 2001;

Jenkins, 1986; Walsh, 2009). Much of the research on effects of the HAB has been atheoretical in nature, in that scholars have gathered data and performed statistical analyses without using a theory to drive the investigation. Although the information gathered is important, it may have less significance because it is data-driven rather than theory-focused. Many early attempts to observe and understand the HAB were anecdotal and did not rely on a theory through which to understand the connection between person and pet.

Scholars and researchers have begun to use terms from attachment theory in efforts to capture aspects of the HAB (Cohen, 2002; Marks, Keopke, & Bradley, 1994; Sable, 2004;

Yorke, 2010). Most of these efforts were not systematic and the authors did not rely on basic tenets of attachment theory when constructing scales of “pet attachment.” Early attempts to operationalize pet attachment did not use the components of separation anxiety, proximity seeking, safe haven, or secure base. Attachment styles and orientations were not used, and human manifestations of the attachment system were often assumed to be equivalent across species (Cohen, 2002; Crawford et al., 2006). Although Bowlby (1970) used animal behavior and ethology in his conceptualization of the human infant-caregiver dyad, he did not indicate that his theory of attachment could be generalized to all animal relationships.

Attachment theory may be an effective means through which to understand the human- animal bond (Zilcha-Mano, Mikulincer, & Shaver, 2011). Before doing this, however, one must gain an understanding of the principal components of Bowlby’s attachment theory. It is also

2

important to examine the evidence that human attachment bonds and figures are meaningful

concepts in relationships across the lifespan before assuming that the theory can be transferred to

a pet-owner relationship. Major components of this theory cite the need for two individuals to:

have an with one another; seek one another in times of distress (safe haven);

yearn for closeness (proximity-seeking); view the other as a base from which to explore

surroundings (secure base); and feel distress when not in contact with one another (separation

anxiety). These components have been explored by multiple researchers over the past fifty years,

and evidence of this bond has been found in observational studies, neurological testing, self-

report measures, and interviews with adults in relationships (Brennan et al., 1998; Hughes et al.,

2012; Schore, 2001; Siegel, 2001). An empirical application of attachment theory to other

constructs, such as the connection between companion animals and their owners allows for a

more expansive investigation of attachment in terms of attachment avoidance and attachment

anxiety. Zilcha-Mano, Mikulincer, and Shaver (2011) have successfully applied attachment

theory to the HAB in Israel, and their self-report measure of pet attachment is now available in

English. The Pet Attachment Questionnaire (PAQ, Zilcha-Mano, Mikulincer, & Shaver, 2011) assesses the HAB in terms of avoidant and anxious pet attachments and provides a reliable measure through which to understand a human’s connection to his or her pet. These researchers have also found experimental evidence that pet owners view their connections with their pets as attachment bonds, and they consider their pets to be safe havens in times of psychological distress and secure bases from which to explore (Zilcha-Mano, Mikulincer, & Shaver, 2011).

Gaps in the Research

A number of shortcomings are present in the current research regarding the connection between pets and their human owners. Most published studies investigating the effects of the

3

human-animal bond are atheoretical in nature, and provide evidence of physiological and

psychosocial benefits of pet ownership without a unifying framework for understanding the

connection. A wide array of results is reported, but there is little to explain the cause,

mechanism, or pathway of the HAB. As a result, one cannot draw conclusions as to the way in

which someone can increase the benefits reaped from pet ownership in a theory-driven,

meaningful way.

Other researchers have used terms from attachment theory to conceptualize the HAB, but they did not formally apply Bowlby’s conceptualization of attachment in the terminology or in measuring the HAB. The resulting studies of “pet attachment” do little to inform scholarly research beyond providing additional evidence that something positive may exist in the bond between human and pet (Antonacopoulos & Pychyl, 2010; Berryman et al., 1985; Brown, 2003;

Geist, 2011). Additionally, there has been a lack of focus on the role of moderating or mediating variables such as pet attachment in better understanding the relationship between adult attachment, daily functioning, and mental health.

The Current Study

More recently, researchers have applied attachment theory as a lens for understanding the

HAB, but their measures have not been explored in a population outside of Israel. The theoretical construct of pet attachment, as defined using an attachment theory framework, has not yet been systematically applied to understand whether pet owners’ connections to their pets are secure, anxious, avoidant, or disorganized. This is a significant distinction because not all connections to

pets are healthy, just as not all attachment bonds are secure. Therefore, this study addressed the

gap in the literature regarding the use of attachment theory as a lens through which to understand

the bond between pet owners and their pets, and the influence of pet attachment on the

4 relationship between adult attachment, social support, and well-being among adults who currently own at least one pet.

Significance

Using an organized theoretical framework of pet attachment as it pertains to perceived social support and subjective well-being could contribute to the literature by adding to what is currently known about the effects of relationship with pets on the owners. As reviews in the field show, little can be said about the general impact of pets on humans. Focusing on individual differences in pet attachment orientation is therefore highly important. This research can contribute to our understanding of the unique contribution of pet attachment to people's well- being and perceived social support, above and beyond interpersonal attachment. It may also provide information which contributes to our understanding of the safe haven and secure base pets provide to people, as reflected in adults’ subjective well-being.

Purpose and Research Questions

The current study investigated the association between pet attachment (PAtt), the relationship between levels of perceived social support (PSS), adult attachment (AAtt), and subjective well-being (SWB) among individuals. In the current study, pet attachment was operationally defined as the depth of attachment anxiety and attachment avoidance within the bond between a companion animal and the human owner. This study addressed several research questions. In what ways are adult attachment style, subjective well-being, and perceived social support related to one another? In what ways are pet attachment and adult attachment related to one another in individuals? How is pet attachment associated with ratings of subjective well- being? Does perceived social support have an association with pet attachment style? Does pet

5

attachment moderate the relationship between subjective well-being and perceived social support?

Hypotheses

Based on the aforementioned research questions, the following hypotheses were tested:

1. Consistent with previous literature, higher levels of secure adult attachment (measured as low

anxiety and low avoidance) will be associated with greater subjective well-being and

perceived social support. More specifically: lower levels of adult attachment anxiety will be

associated with greater subjective well-being; lower levels of adult attachment avoidance will

be associated with greater subjective well-being; lower levels of adult attachment anxiety

will be associated with greater perceived social support; and lower levels of adult attachment

avoidance will be associated with greater perceived social support.

2. Consistent with previous literature, perceived social support will be positively correlated with

subjective well-being.

3. Consistent with previous literature, perceived social support will positively predict subjective

well-being.

4. Adult attachment security will predict subjective well-being.

5. Perceived social support will predict adult attachment.

6. The association between perceived social support and subjective well-being will be partially

mediated by adult attachment.

7. Lower levels of perceived social support will strengthen the association between adult

attachment and subjective well-being.

The following exploratory hypotheses were tested:

6

8. Pet attachment security will be positively correlated with adult attachment, such that pet

attachment anxiety will be positively correlated with adult attachment anxiety; and pet

attachment avoidance will be positively correlated with adult attachment avoidance.

9. Similar to adult attachment findings, higher levels of secure pet attachment (defined as low

anxiety and low avoidance) will be correlated with greater subjective well-being and

perceived social support. More specifically: lower levels of pet attachment anxiety will be

associated with greater subjective well-being; lower levels of pet attachment avoidance will

be associated with greater subjective well-being; lower levels of pet attachment anxiety will

be associated with greater perceived social support; and lower levels of pet attachment

avoidance will be associated with greater perceived social support.

10. Pet attachment security will predict subjective well-being after controlling for adult

attachment.

11. Perceived social support will predict pet attachment after controlling for adult attachment.

12. The association between perceived social support and subjective well-being will be partially

mediated by pet attachment.

13. Lower levels of perceived social support will moderate the association between pet

attachment and subjective well-being after controlling for adult attachment.

7

CHAPTER TWO

Literature Review

In reviewing the literature in this area, one must first explore the theoretical and empirical aspects of attachment theory and the human-animal bond (HAB), and then closely examine and critique more recent attempts to treat them as interchangeable concepts. Once these concepts have been defined, examined, and understood, one can build an argument that attachment theory may be an appropriate lens through which to conceptualize the bond between humans and their pets. Individuals’ attachment to their companion animal may be of clinical significance when examining their attachment style.

Attachment Theory

Tenets of Attachment Theory

In recent years, some researchers and scholars have described the HAB as a form of attachment. Before making such an assumption, however, it is important to understand the theoretical underpinnings of attachment as it relates to humans and the caregiver-infant relationship. Much has been published on the neurobiological or physiological aspects of attachment in humans (Beatson & Taryan, 2003; Fonagy et al., 2011; Schore, 2001; Siegel,

2001), as well as health outcomes based on attachment style or orientation (e.g., Hughes et al.,

2012; Lipton & Fosha, 2011; Sadava et al., 2009). A number of measures also exist for quantifying and classifying attachment style (e.g., Sperling et al., 1996; Strathearn, 2011). Once these concepts are understood, one can then decide whether and how to apply attachment theory to the HAB (e.g., Noonan, 2008; Sable, 2004).

The originator of attachment theory, was trained as a psychoanalyst, but his work was rejected by the analytic community, which viewed it as “superficial in its focus on the

8

importance of real-life events such as separations and losses rather than on the primacy of

internal fantasy” (Connors, 2011, p. 348). Unlike the theory used by analysts of his time,

Bowlby’s theory is relational in nature and focuses on , views of the self, views of

others, and interpersonal interactions. In his theory of attachment, Bowlby used concepts from

ethology, evolutionary biology, cognitive psychology, object relations, and systems theory. He

emphasized the critical role of attachment and connection in the development of personality and

overall functioning. Bowlby rejected Freud's drive theory in favor of a more ethological outlook,

noting that infants and children instinctively seek closeness and attachment to caregivers for the

purposes of protection, safety, and nurturing. Bowlby referred to this as a control systems

approach and likened it to the purpose and utility of a thermostat. When a room is too cold, for

example, the thermostat senses this and activates the heater in an effort to regulate the

temperature of the room. Once the desired temperature is attained, the thermostat again senses this and turns off the heater. In a similar way, Bowlby hypothesized that children need to be physically close to their caregivers. If a physical separation becomes too great, for example, the attachment system is activated by way of fear, prompting the child and/or caregiver to seek

proximity to one another in order to feel safe and reduce the perception of threat. Once sufficient proximity is achieved, fear is reduced and the system achieves homeostasis (Bowlby, 1982).

It is important to understand that the attachment behavior system is the organization of

attachment behaviors within a person. Attachment behaviors serve an individual’s goal of

promoting closeness to an attachment figure (Cassidy, 1999). Bowlby identified four primary

attachment behaviors: proximity seeking, separation anxiety, safe haven, and secure base (see

Figure 1). Proximity seeking and separation anxiety may be explained by the aforementioned

thermostat example. When an individual believes the attachment figure is physically or

9 emotionally unavailable, the individual may experience fear or anxiety and seek ways to re- establish closeness to the attachment figure. A third type of attachment behavior is safe haven, which acknowledges that an individual will seek his or her attachment figure in times of distress as a source of comfort and affection. Secure base is the fourth form of attachment behavior. This occurs when an environment is judged to be safe to explore, the attachment figure is perceived to be physically or emotionally available, and the individual feels comfortable exploring his or her environment. Attachment behavior is closely connected to fear response, in that both serve to protect the individual (Cassidy, 1999). The attachment behavioral system serves a number of important components. It has an evolutionary function because it enhances the likelihood of survival and continued existence of the species. The system is also teleological or goal-oriented in nature, and attachment behaviors (proximity-seeking, safe haven, secure base, and separation anxiety) are utilized in an effort to protect one from danger, injury, or demoralization

(Mikulincer & Shaver, 2007).

Figure 1: Components of the Human Attachment Behavior System

proximity seeking

Attachment separation safe haven anxiety Behavior System

secure base

10

Figure 1. As identified by Bowlby (1982) attachment between and infant and a caregiver is characterized by the presence of identified four primary attachment behaviors: proximity seeking, separation anxiety, safe haven, and secure base.

In addition to understanding the attachment behavior system, one must also consider the importance of an attachment bond. Cassidy (1999) defined it as “the bond that one individual has to another individual who is perceived as stronger [or] wiser” (p. 12). An attachment bond is an extension of an affectional bond, which possesses the following qualities: it is persistent as opposed to transitory; it is a specific connection from one individual to another; it is part of an emotionally significant relationship; it strives for proximity to the other person; and anxiety is elicited when the two individuals are separated. The difference between an attachment bond and an affectional bond is that in addition to possessing the components of an affectional bond, in an attachment bond, the individual seeks security and comfort in the relationship. When the individual seeks and achieves this security, the attachment is considered secure; when the individual seeks but does not attain this security, the attachment is considered insecure. The rupture of an attachment can lead to emotional distress and maladaptive methods of getting one's needs met. In reviewing major tenets of attachment theory, Sable (2004) noted its relational, spatial, and affective components. In addition to exploring the meaning of the bond, one must also account for the role of physical proximity or closeness, as well as account for how the individual experiences and responds to .

In understanding attachment theory, one must also consider attachment style. Attachment styles refer to patterns of expectations, needs, feelings, and behaviors derived from an individual’s experiences in relationships, particularly the infant-caregiver relationship. These dimensions are also relevant in adult relationships. In more recent years, scholars have identified

11

anxiety and avoidance as being of primary importance in understanding adult attachment styles

(Mikulincer & Shaver, 2007). Attachment-related anxiety is often conceptualized as a cluster of

associated behaviors, thoughts and feelings: yearning for closeness and connection (proximity- seeking), rumination and worry about the attachment figure’s availability, concern about how one is valued by the attachment figure, and the employment of approach behaviors in an effort to reduce distress. Attachment-related avoidance is characterized by a different pattern: discomfort and aversion to , a desire to remain emotionally distant and independent from an attachment figure, and the use of disengagement strategies to reduce distress (Mikulincer &

Shaver, 2007) (See Figure 2 for additional information).

Figure 2: Additional Concepts Within Attachment Theory

Attachment Attachment Attachment Affectional Figure Style Behaviors Bond

proximity secure persistent over person to whom seeking one is time affectionally bonded separation disorganized anxiety specific between two secure people insecure-avoidant base sought when wanting security emotionally insecure- anxious safe haven significant

Figure 2. Components within attachment theory.

12

As Mikulincer and Shaver(2007) noted, “Although the central concepts and tenets of attachment theory are fairly easily to understand, it is important not to equate them too readily with everyday conceptions of human motivation and relationships” (p. 17). Attachment bonds are specific to two individuals, and the meaning of an attachment figure is quite specific.

Attachment figures are more than relationship partners. When a person is experiencing distress and is in need of support, they are the individuals sought for the provision of safe haven, secure base, and physical or emotional closeness. In adulthood, attachment figures may be symbolic representations or actual individuals who are sought for their abilities to provide safety, protection, and comfort.

Attachment theory is applicable across the lifespan, and attachment systems are flexible and adaptable over time. In infancy, for example, physical proximity and nurturing may be the most salient goal. As a result, the behavioral strategies associated with an infant in distress

(crying or reaching out to the attachment figure) seem to be innately targeted to alert the caregiver to the infant’s discomfort. In secure attachment, the caregiver responds to the infant in ways which allay the child’s fear and reduce distress. As an individual matures and develops, however, the attachment system becomes more context-specific and flexible. For adults, the primary strategy may not include seeking physical proximity. An adult’s need for safety and security in times of distress may be met by mental representations or cognitive priming of an attachment figure. For example, an adult may telephone an attachment figure as opposed to needing physical touch or contact with the person (Mikulincer & Shaver, 2007).

Measuring and Quantifying Attachment Style and Orientation

As Strathearn (2011) noted, the two primary measures of attachment are the Ainsworth

Strange Situation Procedure (ASSP) and the Adult Attachment Interview (AAI). In the ASSP, an

13

infant and his or her caregiver are placed in a novel environment and introduced to an unfamiliar

adult, then experience three episodes of separation and two episodes of reunion. The instances of

separation are intended to activate the infant’s attachment system, meaning that the infant

becomes distressed or fearful, and when reunification takes place, the infant seeks proximity and

comfort with his or her caregiver. Eight total situations are observed: the and infant are

introduced to the room; the parent does not interact with the infant while the infant is allowed to

explore the room; a enters the room, talks with the parent, approaches the infant, and the

parent inconspicuously leaves; the stranger approaches the infant; the parent returns, comforts

the infant, and leaves the room with the stranger; the infant is left alone in the room; the stranger

enters and approaches the infants; and finally, the parent enters the room, greets and picks up the

infant, and the stranger leaves the room inconspicuously. Throughout the eight situations, four

aspects of the child’s behavior are measured: the infant’s level of exploration in playing with

new toys (assessing secure base); the infant’s reactions to the parent’s departure (assessing

separation anxiety); the infant’s reaction to the stranger (assessing stranger anxiety); and the infant’s behavior when reunited with the parent (assessing safe haven and proximity-seeking)

(Ainsworth et al., 1978).

On the basis of their observed behaviors, infants’ attachment styles could then be understood as secure, anxious-resistant insecure, anxious-avoidant insecure, or disorganized

(Ainsworth et al., 1978; Main & Solomon, 1990). An infant exhibiting a secure attachment style will explore freely and interact with the stranger while the parent is present, will be visibly distressed when the parent leaves the room, and will exhibit happiness when the caregiver returns to the room. When the parent is out of the room, the securely attached child will not interact with the stranger. The style has been characterized as the most adaptive connection

14

between child and attachment figure, and Ainsworth (1978) suggests that this attachment comes

when the child perceives the parent as present and able to meet his or her needs in a responsive

and adaptive manner. If the infant has an anxious-resistant insecure attachment with the caregiver, the child will appear anxious of exploration and interaction with regardless of the parent’s presence. When the parent leaves the room, the anxious-resistant child becomes extremely distraught, and the reappearance of the caregiver is met with ambivalent emotions and behaviors, as though the child wants to be physically close to the parent but harbors resentment.

This form of attachment has been associated with a parenting style that is itself inconsistent, in

that the child’s needs are sometimes met and sometimes ignored. In children with anxious-

avoidant insecure attachment, behaviors are more disengaged and lack emotion. The parent and

stranger may be treated similarly, in that both adults are largely ignored or disregarded, and little

emotion is observed when the adults enter or leave the room. Similarly, the infant does not

engage in exploration or proximity-seeking behaviors. This form of attachment has been

associated with caregivers who do not engage regularly with the child, with the result that the

child’s needs are rarely met and the child comes to believe that communication with the parent is

unimportant (Ainsworth et al., 1978). The fourth attachment style which may be observed in the

ASSP is disorganized or disoriented. In this category, which was often observed in parent-child

dyads in which the parent experienced trauma or loss just before or after the birth of the child,

the child was observed to cry during separation, but then avoided the caregiver upon reunion. In

other instances, the child may approach the caregiver upon reunion, but then freeze or fall to the

floor (Main & Solomon, 1990).

In an effort to clearly distinguish the two and quantify adult attachment styles, the AAI is

transcribed and analyzed to illuminate patterns of behavior and relationship quality, which then

15

allows the romantic partner’s attachment pattern to be considered secure, insecure⁄ dismissing, or insecure⁄ preoccupied. In these styles, the secure adult attachment was similar to secure attachment as measured by the ASSP, insecure/dismissing adult attachment was similar to anxious-avoidant insecure attachment in the ASSP, and the insecure/preoccupied adult attachment was similar to anxious-resistant insecure attachment in the ASSP. The AAI has been validated through its use in more than 200 published studies, and several trends about the three groups have emerged: dismissing adults tend to rely more heavily on rules and past experiences

than on their own emotions or affective information when making decisions; adults with

preoccupied attachment styles make decisions based on emotional components and their own

feelings, rather than facts or observable behaviors; and adults with secure attachment styles seem

to integrate affective and cause-effect information in a balanced way when making decisions

(Strathearn, 2011).

Acknowledging that the ASSP and the AAI are considered the most reliable indicators of

attachment style according to theory (Strathearn, 2011), one must also acknowledge the time-

intensive nature of administering, coding, and interpreting both assessments. In light of this,

numerous self-reports have emerged over the past 30 years. Roisman et al. (2007) further

clarified the difficulty in assessing attachment style in their review and analysis of published

measures, finding that only a small degree of overlap exists between self-reported attachment

style measures and the Adult Attachment Interview’s conceptualization of secure attachment.

They found that self-reports were consistent with individuals’ appraisals of the emotionality in

their relationships, while the results of the Adult Attachment Interview were only trivially

correlated with self-reports of relationship functioning. As such, the minimal overlap between

the two forms of attachment measurement suggests that results from the AAI and self-report

16 literatures should not be cited and discussed in narrative reviews as if the measures were interchangeable. Their suggestion to ameliorate this discrepancy would be for future researchers to focus on the role of attachment-related threat, as observed in the Ainsworth Strange Situation

Procedure (ASSP) or by asking more specific questions which clarify the specific component of attachment (Roisman et al., 2007).

As previously stated, researchers have discovered two dimensions of attachment insecurity in self-report measures: anxiety and avoidance. One such measure, constructed by

Brennan, Clark, and Shaver (1998), is the Experiences in Close Relationships Scale (ECR). The authors factor-analyzed the items in numerous published self-report scales and synthesized the results with coding scales from the ASSP. The resulting measure was comprised of two 18-item scales, one to assess attachment anxiety and a second scale to gauge avoidant attachment. The

ECR has been used in hundreds of search studies, and researchers have translated it into multiple languages. Across its use, it has demonstrated high reliability (α > .90 with test-retest coefficients between .50 and .75) and discriminant validity (correlation between anxiety and avoidance subscales is close to zero). The validity of the ECR has been tested across experimental manipulations and behavioral observation. The ECR provides helpful information about whether an individual tends to have an anxious, avoidant, or disorganized/preoccupied attachment style in relationships. One criticism of the assessment is that it defines secure attachment as being the absence of attachment anxiety and avoidance, which could be viewed as a nebulous definition.

Despite this limitation, the ECR and its revision continue to serve as benchmarks for measuring adult attachment style because they utilize formalized attachment theory, build on the ASSP and

AAI coding protocols, and have consistent validities and reliabilities across time and culture. In addition to this, the wording of the ECR has been successfully altered to capture a participant’s

17

global attachment style, his/her attachment style in a specific romantic relationship, and his/her

retrospective attachment style in other relationships (Mikulincer & Shaver, 2007).

Neural Pathways of Attachment

Attachment in children. Attachment has been considered a biologically instinctive process that drives an infant to create and maintain meaningful bonds with a caregiver in an effort to feel secure, explore his or her world, and engage in self-soothing behaviors when distressed (Siegel, 2001). This is critical for helping individuals begin to develop ideas about themselves, as well as what to expect from others. Attachment relationships are crucial to the emotional and social development of a child and continue to be important throughout the lifespan. Siegel (2001) postulates that, in the first years of life, interpersonal experiences with a caregiver may have the broadest impact on the brain structures that mediate social and emotional functioning. Neurons, the basic cells of the brain, receive and transmit messages in the form of electrical activity or action potentials. When an electrical pulse is occurring in a circuit or pathway, it stimulates the release of a chemical known as a neurotransmitter that flows across the small synaptic space between nerve cells, to activate (or inhibit) the receiving neuron. When interconnected neurons, groups of neurons, and clusters of groups of neurons that form specific pathways and systems of the brain are activated, mental processes are produced. Energy and information can flow within one brain, or between brains, in the form of interpersonal communication. The various ways in which impulses and data flow within an individual or between two individuals helps create the experience of mind, and also serve as attachment relationships, which is important to understand in reviewing the literature on attachment theory, the human-animal bond, and well-being (Siegel, 2001).

18

Beatson and Taryan (2003) cited the importance of the right brain hemisphere, noting that

it tends to be dominant in the first three years of life, has a significant role in attachment

behaviors, and plays a dominant part in the processing of emotion. The right hemisphere has

also been implicated in neuroendocrine and autonomic activation, as well as the stress response.

This part of the brain seems to be responsible for the perception of emotion, facial expression of

affect and non-verbal aspects of language, and the ability to conceptualize the mind of another.

Schore (2001) also identified the right hemisphere as the seat of control over parasympathetic

and sympathetic responses to stress, and acknowledged that this area has a significant role in

regulating physiological and hormonal functions throughout the lifespan. The right hemisphere is

dominant for the production of cortisol, corticotrophin releasing factor CRF, and

adrenocorticotropic hormone (ACTH). Additionally, the hypothalamopituitaryadrenocortical

(HPA) axis and the sympatheticadrenomedullary (SAM) axis are controlled by the right cerebral

cortex. Both of these pathways are crucial in helping an individual survive and adapt both

actively and passively to distress (Schore, 2001).

From the period prior to birth and extending through the first three years of life, neurons

and synapses are overproduced, but as a child matures, pruning (selective elimination of the

genetically produced excess) shapes development as neural pathways and circuits are formed. A

different process, sometimes referred to as experience-dependent development, happens when

new neural connections are created as a result of experiences or interactions with the child’s

environment, suggesting that experience has the potential to alter brain structure by leading to

either the maintenance and strengthening of existing synapses, or by the experience-driven creation of new synaptic connections. These neuronal connection patterns determine brain functioning, as well as how the individual will perceive the world and his or her interpersonal

19 interactions. Ongoing experiences may lead to the development of deeper, more intricate neural integration (Siegel, 2001).

In early life, the child’s frontolimbic system responds to social inputs by encoding expectations of safety and security from the caregiver. Across the lifespan, the amygdala identifies emotional stimuli while the hippocampus stores stress-related information into long- term memory. The amygdala also acknowledges signs of threat and functions to regulate stress hormone secretion by the hypothalamus to allow soothing. Feedback loops between the prefrontal cortex, amygdala, hippocampus, and hypothalamus contribute to memory formation and conditioned learning, including the appraisal of and response to emotional stimuli. These associations become strengthened circuits through dopaminergic and oxytocinergic activity in the hypothalamus, nucleus accumbens, ventral tegmentum, and amygdala, and are experienced as emotional and physical security. The development and reinforcement of these circuits shapes interpersonal relationships throughout an individual’s life (Hughes et al., 2012). Schore holds that emotion regulation is more than the reduction of affective intensity because it also involves an intensification of positive emotion. In agreement with Siegel, Schore (2001) observes,

“Attachment is not just the reestablishment of security after a dysregulating experience and a stressful negative state, it is also the interactive amplification of positive affects, as in play states” (p. 21).

Unfortunately, while secure attachment is the goal in childrearing, it is not always the outcome. Frontolimbic dysfunction is common to many psychiatric conditions, and its ties to attachment cannot be overstated. These dysfunctions are manifested as problematic affect, behavior, interpersonal relatedness, and cognitions (Hughes et al., 2012). When the caregiver is a source of stress and danger, or a traumatic event overwhelms the system, effects can be long-

20 lasting and include shrinking of the hippocampus and chronic activation of the amygdala. Long- term symptoms of this chronic stress and sympathetic nervous system (SNS) activation include sleep disturbance, chronic pain, muscle tension, panic, chronic rage, weakness, exhaustion, or concentration difficulties. Another potential outcome could be simultaneous activation of the

SNS and parasympathetic nervous system (PNS) resulting in the dissociative freeze response, typified by activation of the dorsal vagal branch of the PNS. Symptoms here would be characterized by hypoarousal and include muscle weakness, depression, chronic fatigue, and gastro-intestinal symptoms (Lipton & Fosha, 2011).

Siegel (2001) identified five elements of caregiving that could lead to secure . Interestingly, these principles could be effectively transferred to the client-therapist relationship, or to the pet-owner relationship. The first, collaboration, refers to communication which is transparent between an attuned dyad, and may be verbal or nonverbal in nature. This element is remarkably similar to Rogers’ (1980) conceptualization of congruence. The second element, reflective dialogue, refers to the process by which a caregiver can correctly interpret the needs of the child and communicate back in a meaningful way, such that the child feels heard in a meaningful way and can believe that his or her internal experience is important and valid.

Siegel refers to this as mindsight, “The capacity of the mind to create the representation of the mind of others, and of the self” (p. 79). The third element, repair, can teach the child that life is fraught with inevitable misunderstandings and miscommunication, but that these ruptures can be recognized and resolved. The fourth characteristic of secure caregiving is the development of coherent narratives, or the combining of the past, present and future in meaningful ways, with the goal of co-constructing an autobiography that makes sense and gives purpose. The final element, emotional communication, refers to the idea that caregivers have the ability to amplify and share

21

in both positive and negative experiences with a child, such that the child can learn that emotions

are acceptable, natural, and meaningful (Siegel, 2001). These components will be important in understanding the potential use of attachment theory and its potential utility in understanding the

HAB.

Strathearn (2011) compared 15 with secure attachment and 15 with insecure⁄

dismissing styles, as categorized by the Adult Attachment Interview (AAI). The groups were

congruent in the areas of socioeconomic status, psychopathology risk, IQ, race or self-reported parenting stress. In functional magnetic resonance imaging (fMRI), however, mothers in the insecure⁄dismissing group expressed a significantly lower level of activation in the ventral striatum and medial prefrontal cortex bilaterally when viewing their own infants’ happy faces, suggesting a lower sense of reward or pleasure from seeing their child experiencing joy. These mothers also exhibited a lower level of peripheral and central oxytocin production when interacting with their infants, possibly indicating a less secure attachment to their children

(Strathearn, 2011). As will be explored in the review of literature in the HAB domain, oxytocin also has an important connection to the HAB.

Attachment in adults and across the lifespan. Fonagy, Luyten and Strathearan (2011) cited physiological evidence for Bowlby’s attachment theory of the secure-base system and the caregiving system, associating them with the activation of the oxytocinergic neural system, which has consistently been shown to play an important role in promoting and maintaining maternal behavior. Oxytocin is a neuroactive hormone that is made in the hypothalamus and projects to brain areas that are connected to emotions and social behaviors (the striatum, amygdala, and cingulate cortex). Vasopressin, a neuropeptide, may play an analogous role to oxytocin, perhaps especially in males, mediating some of the dysregulated social experience of

22

disorganized attachment. These chemicals have roles in two aspects of creating attachment

relationships, by activating the reward/attachment system, and by deactivating neurobehavioral

systems that are involved in mediating social avoidance. Oxytocin is traditionally known for its

part in female reproduction, as it is released during childbirth and breastfeeding. Oxytocin

concentrations during early pregnancy and during the postpartum period are strongly associated

with maternal bonding as suggested by indicators such as positive affect, gaze, affectionate

touch, and frequency of monitoring the infant as well as thoughts related to attachment. Studies

with intranasally administered oxytocin have shown oxytocin to improve social memory,

increase trust, generosity, and the experience of empathy associated with watching emotional

videos in humans (Fonagy et al., 2011).

Oxytocin facilitates social connectedness at a number of levels. It has been found to

facilitate social function in experimental studies, including improving social memory, and

memory of facial expressions and identity, and appears to selectively facilitate positive, happy

social memories. It creates or activates a positive emotional experience in relation to other

people, and seems to generate a sense of trust in interpersonal relationships. Oxytocin lessens

behavioral and hormonal responses to social stress. In animal research it has also been shown to animals to overcome their natural avoidance of proximity and to inhibit defensive behavior, thereby facilitating approach behavior (Vasopressin has primarily been implicated in male- typical social behaviors, including aggression and pair-bond formation, and mediates anxiogenic effects. Oxytocin can inhibit hypothalamopituitary stress axis activity, suggesting that activation of the attachment system may generate increased experience of reward, increased sensitivity to social cues, and decreased social avoidance, but also the potential for the reward to override lack of trust (Fonagy, Luyten & Strathearan, 2011). Riedl and Javor (2012) discussed the role of

23 arginine vasopressin (AVP) in interpersonal communication and behavior. The duo hypothesized it to be associated with avoidance and distrust because it is connected to aggression, territoriality, and heightened response to stress. These constructs are important in understanding the roles of attachment anxiety and attachment avoidance, particularly as they can be manifested in adult attachment relationships (Riedl & Javor, 2012).

Strathearn (2011) also identified oxytocin as important in adult attachment, noting that it has the ability to provide anxiolytic and bonding effects. His research focused specifically on the maternal bond and its role in cases of physical or emotional neglect, and traced several attachment circuits in the brain. In randomized, placebo-controlled experiments, the introduction of intranasal oxytocin produces a broad range of pro-social effects, including enhanced social memory, increased eye contact when looking at faces, enhanced recognition and memory of facial expressions and identity, and increased perception of trust. The neuropeptide hormone oxytocin is synthesized in the hypothalamus and released into the bloodstream or into specific brain regions. In the amygdala, it has an anxiolytic effect and facilitates social recognition, while in the hippocampus, it aids in the development and maintenance of long-term memory. Just as the presence of oxytocin is important, its absence or deficit is also meaningful. Women who survived emotional neglect and other, nonphysical form of maltreatment in childhood have significantly reduced concentrations of oxytocin in their cerebrospinal fluid. A positive correlation exists between mother and infant salivary oxytocin levels (Strathearn, 2011).

As Schore (2001) noted, the stress response is mediated by the sympathetic– adrenomedullary (SAM) axis and the hypothalamo–pituitary–adrenocortical (HPA) axis. The threshold for stimulation of the SAM axis is lower than that for stimulation of the HPA axis, and is regulated by corticotrophin releasing factor (CRF), which regulates catecholamine release in

24

the sympathetic nervous system. The HPA axis is regulated by cortisol, the major antistress

hormone. As the severity of the stressor increases, more neurochemicals are produced and

released. These systems function synergistically. Catecholamines facilitate the availability of

energy to the body’s vital organs, while cortisol inhibits activation of the sympathetic nervous

system. In other words, the energy-expending sympathetic and energy-conserving

parasympathetic components of the autonomic nervous system (ANS) regulate the physiological

and psychological experience of stress, which is an experience which may trigger attachment

avoidance or anxiety in adult relationships (Schore, 2001).

Beatson and Taryan (2003) extended Siegel’s discussion on attachment and brain development, focusing on the etiology of depression. They acknowledged the role of stress on the developing brain but also discussed the role of neuroplasticity in ameliorating the effects of stress or maladaptive attachment. Neuroplasticity refers to the brain’s ability to integrate novel experiences and modify genetic expression, thus altering the functioning of neural circuits. Such alterations can have a significant impact on attachment behaviors, in addition to other experiences. It is important to note, however, that these pathways can be changed or adapted over time with counseling, reparative experiences, or the creation of a secure attachment bond with an attachment figure (Beatson & Taryan, 2003).

At the physiological level, Coan (2010) noted that adult attachment relationships rely on the substances vasopressin and oxytocin, the latter of which has been associated with the development and maintenance of trust. According to Coan, attachment styles arise “past experiences combining potential threats with the presence, absence, or specific behaviors, of early attachment figures… [and] manifest as two independent axes: attachment anxiety and attachment-related avoidance” (p. 212). Despite advances in brain imaging technology,

25

attachment style has not been observed or traced in the brain. One may observe their effects,

such as asymmetries in functioning or the activation of specific pathways known to be associated

with affect and behavior. According to Connors (2011), insecure attachment affects

corticolimbic functioning, the area of the brain involved in emotion regulation. In persons with a

dismissing attachment style, emotional processing is blocked in the right brain hemisphere.

Preoccupied attachment is associated with overexpression of episodic memory in the right brain

hemisphere. Attachment styles tend to be consistent over time, and also have the capacity to

change in response to new experiences, such as therapeutic encounters or healing relationships

with people or companion animals (Connors, 2011).

Sadava and his colleagues (2009) sought to identify connections between attachment and

overall health. They theorized that attachment orientation could impact an individual’s health in

multiple ways: positive and negative emotional or affective states, responses to stress, perceived

social support, and health or risk behavior (such as substance use). They conducted a path

analysis of these four linkages to determine whether attachment avoidance or anxiety could

indirectly impact physical health via affect, stress, social support, and/or health-risk behavior.

They hypothesized higher levels of attachment anxiety and avoidance would be related to greater

negative emotionality, less positive emotionality, higher perceived stress, lower perceived social

support, and more frequency of behaviors deemed risky. They further hypothesized that these

factors would have a deleterious effect on an individual’s physical health. They sampled two

different groups: 623 first-year undergraduate students and 219 clients completing an initial

intake for addiction treatment (Sadava et al., 2009).

The participants completed the 30-item Relationship Scales Questionnaire (RSQ; Griffin and Bartholomew, 1994) to assess relationship style, the 20-item Positive and Negative Affect

26

Schedule (PANAS) to assess emotionality, and the Social Support Questionnaire (SSQ) to measure perceived social support. They also answered questions about current stress level, alcohol use, marijuana use, and tobacco use to determine overall stress and health-risk behavior information. Using structural equation modeling, the researchers found an indirect connection between attachment orientation and physical health. In the undergraduate student sample, attachment orientation was significantly correlated with affect, stress, and social support.

Additionally, attachment anxiety was linked with smoking. In the addiction sample, attachment style was significantly correlated with affect and stress and attachment anxiety was connected with negative emotionality, stress, and marijuana use. In examining the overall sample, affect and stress significantly predicted health outcomes (Sadava et al., 2009). This research, while expansive in nature, is not without limitations. The study was cross-sectional in design, which prevented a causal association from being made. Additionally, the use of researcher-created questions, as opposed to clinically-derived and validated scales, weakened the findings about stress and health-risk behavior. As such, it would be difficult to generalize the findings about these areas without further review of the specific questions asked. Both samples were reported to be in the midst of significant transitions, and the researchers assumed that overall stress level would be greater for participants, but this was not specifically assessed or controlled with a more stable, non-transitory but demographically similar sample. Each measure used in the study relied on participant self-report. It would have been advantageous to include more objective or standardized measures (Sadava et al., 2009). Despite the limitations of the research, the finding that attachment orientation has an indirect connection to physical well-being is of significance to the current study, as is the use of the PANAS to assess subjective well-being.

27

Mikulincer and Shaver (2007) cited the benefits of having an attachment figure who was perceived as being physically available and supportive and acknowledged that this connection can lead to a significant reduction in distress, as well as an increase in overall positive affect. The physical presence of a close friend or romantic partner during a stressful experience (performing arithmetic, threat of an electric shock, or in interpersonal interactions) was associated with a reduction in blood pressure, skin conductance, and stress response (as measured by functional magnetic resonance imaging). The presence of the attachment figure had a buffering effect on the stress response, such that the participant’s arousal was lower when with a friend than when with the experimenter in a laboratory setting (Coan, Schaefer, & Davidson, 2006; Edens, Larkin,

& Abel, 1992; Kamarck, Manuck, & Jennings, 1990). Gump and colleagues (2001) found that participants’ blood pressure was lower when interacting with a romantic partner than when alone or interacting with other people, suggesting that the mere presence of the attachment figure leads to a reduction in physiological distress, even in a non-laboratory setting.

Evidence also exists that attachment relationships can provide extensive social support

(Mikulincer & Shaver, 2007). Perceived social support from a romantic partner has been consistently associated with an increase in one’s perceived ability to cope with stress, reducing anxiety and rumination, and a significant reduction in one’s experience of pain (Cohen &

McKay, 1984; MacDonald & Leary, 2005; Thoits, 1986). Additionally, in assessing couples and close using self-report measures, greater levels of perceived social support from the attachment figure were significantly correlated with enhanced feelings of safety, relationship satisfaction, and overall positive affect (Deci et al., 2006). Within the realm of secure attachment orientations, researchers have found consistent links between relationship security, perceived social support, and elevated mood (enhanced well-being). In adults, even the symbolic

28

availability of attachment figures, through priming of attachment figure representations or asking

a participant to visualize an attachment figure, has a soothing effect (Mikulincer & Shaver,

2007).

The reverse of theses also appears to be true. Finch and colleagues (1999) completed a meta-analysis of 48 published studies and concluded that negative or distressing interactions with romantic partners negatively impacted subjective well-being. These unpleasant interactions could be perceived criticism or rejection from a romantic partner, and effects included increased levels of depressed mood, anger, and anxiety in participants (Finch et al., 1999). In measuring the pain response, MacDonald and Leary (2005) found that rejection from romantic partners or the experience of unrequited love can lead to activation of brain areas which are associated with detecting physical pain in adolescents.

Animal models and bonding. In more recent years, animal studies of bonding have found that physical closeness and communication within the infant-mother dyad provide physical protection as well as have significant regulatory impacts on the behavioral and physiological functioning of the pair. Insel and Young (2001) cited mother-infant animal models of attachment, such as sheep, voles, and rats, to help explain some of the processes in the human brain. Recent advances and studies utilizing positron emission tomography (PET) scans and functional magnetic resonance imaging (fMRI) have allowed neurobiologists to identify physical evidence of attachment networks in the brain, which “are a consequence of nurturing or lack thereof, by an attachment figure” (Yorke, 2010, p. 562). As previously noted, attachment networks and pathways are present in the brain. A related concept which has not yet garnered physiological evidence is neural resonance, which involves mirror neurons. Yorke hypothesized that this attunement in affect can occur between two living things, suggesting that such an

29

emotional current could move beyond copying or mirroring the actions of an attachment figure.

Further, the presence of neural resonance could “facilitate interactions in which individuals often perform complementary actions to achieve a common goal” (p. 562-3). Evidence for the role of mirror neurons has been established with primates and lends credence to the concept of empathy being able to exist between two beings. Attunement and resonance have invaluable roles in the therapeutic relationship, and Yorke's conclusions suggest that these constructs may exist between animals and humans (2010).

Benefits of the Human-Animal Bond (HAB)

Multiple studies have suggested that pets can offer a therapeutic function and this is

sensible, given the human need for connection and the historical significance of a meaningful

bond between human and pet (Cavanaugh, Leonard, & Scammon, 2008; Morrison, 2007;

Ormerod et al., 2005; Ovesen, 2005; Walsh, 2009). When considering the role that pets play in

the life of a human, the concepts of unconditional positive regard and congruence seem ever-

present (Rogers, 1961). Empathy, while characterized as a distinctly human trait, also has been

observed in service animals that are often able to forecast and prepare for the needs of their

owners by allaying anxiety or alerting to the presence of an oncoming epileptic seizure (Noonan,

2008; Walsh, 2009).

In recent decades, the emergence of animal-assisted interventions and their potential role

in healing has exploded, with thousands of articles being published on a wide array of animal-

assisted interventions (literature reviews and meta-analyses include Barker, 2005; Beck, 1985;

Beck et al., 1986; Crawford et al., 2006; Draper et al., 1990). Pets have been introduced to assisted living facilities, served as mascots in nursing homes, been used in play therapy with

children, been raised in correctional facilities, and can now become certified service animals for

30 varying purposes, assisting those with physical disabilities and mental illnesses alike. Pets have also been introduced to the psychotherapy environment, most often with the therapist introducing his or her own pet to the counseling atmosphere. Many articles provide anecdotal evidence of the human-animal bond, or the use of pet therapy, but lack empirical support for claims of increased well-being or positive change in the lives of humans. Cangelosi and Embrey (2006) chronicled the life and use of a pet dog in an assisted living facility, and quoted the dog: “I admit I love the attention, but I derive great satisfaction from comforting someone who is lonely, and I love to help someone remember happy times” (p. 19). Tilsen (1998) detailed several vignettes of using dogs in therapy. Imber-Black (2009) offered a similar account of Snuggles, her pet cat, as a “co- therapist” in sessions. Noonan (2009) drew on attachment theory and the process of grief in describing the death of her pet cat, and she opened by acknowledging, “This paper arises not out of professional practice, but out of a personal experience and the subsequent reading and conversations with professional colleagues and friends” (p. 395). Such published pieces serve the purpose of drawing in the reader, but lack empirical support or clinical significance, which is needed in order for the field to gain legitimacy.

Other studies that discuss the human-animal bond are more focused on implementation of pet therapy, and provide anecdotal rationales for the use of pets, followed by steps or suggestions to beginning a program in general (Wilson & Barker, 2003) or in more specific settings, such as in hospitals (McCandless, McCready, & Knight, 1985; Yamauchi et al., 1984), assisted living facilities (Davis & Juhasz, 1984; Kovacs et al., 2004), when working with children (Esteves &

Stokes, 2008; Friesen, 2010; Walsh, 2009), hospices (Brodie & Biley, 1999; Chinner & Dalziel,

1991), when elderly patients (Jessee, 1982; Michaels, 1982; Ryder, 1985; Wilson &

Netting, 1983), or in the field of social work (Risley-Curtiss, 2010). Barker (2005)

31

acknowledged the role of companion animals throughout multiple domains, and she observed the

increasing use of pets in advertising products for human consumption and a growing trend of

employers allowing pets in the workplace. While interesting and colorful, such accounts lack the

rigor, theory, or methodology needed to justify the application of such interventions.

Important Terminology in the Human-Animal Bond

Prior to exploring the human-animal bond (HAB), it is important to understand the various terms used within the literature, as well as to appreciate the focus of the current study in examining the bond between a pet owner and his or her pet. Anthrozoology refers to the study of the interaction between humans (anthropos) and animals (zoon). Some researchers refer to this as human-animal interaction (HAI). As a result, the terms anthrozoology and HAB are interchangeable (http://www.anthrozoology.org/). Within the scope of HAI lie animal-assisted

interventions (AAIs), which are activities or experiences which feature an animal as part of an

intervention, and the human-animal bond (HAB), which focuses on the affectional connection

between humans and animals. Within AAIs and the HAB lie additional specifications so it may

be helpful to review Figure 3 to understand which aspect of HAI is being explored. In this

review, the author reported on the physiological and psychosocial benefits of the HAB. These

effects were often reported as outcomes of AAIs. In some studies, however, these benefits were

reported as an outcome of pet ownership, which was not considered to be an AAI (see Figure 3).

32

Figure 3. Forms of Human-Animal Interaction.

Figure 3. There are multiple ways in which to classify and define the connection between humans and animals.

Many authors of the studies reviewed use the term companion animal. It may also be helpful to use the following illustration to understand how authors could classify an animal.

Again, this study is specifically focused on the experience of pet ownership (See Figure 4).

Figure 4. Classifications of Animals.

Animals

Wildlife Domesticated Animals

Livestock Companion Animals

Service Therapy Pets Animals Animals

Figure 4. The current study will focus on pets.

33

Evidence of the Physiological Benefits of the HAB

As previously noted, humans have chosen to keep and care for pets for thousands of

years, and they have treated cats and dogs as servants, guardians, and companions (Walsh,

2009). Given the historical connection between person and pet, researchers have sought to

measure the physiological effects of the human-animal bond (HAB). In many cases, they have found evidence that the HAB has physiological benefits for both the human and the companion animal (e.g., Astrup, Gantt, & Stephens, 1979; Cole et al., 2007; Odenaal & Lehmann, 2006).

When considering the interaction and meaning of the HAB it is important to understand the biochemical and physiological processes which occur in the establishment of the HAB. One must also acknowledge the outcomes of these processes. In a general review of literature pertaining to the HAB, one may find articles encompassing a wide array of topics. Beck and Katcher (1984) offered a critique regarding the reliability and accuracy of results reported in early studies,

observing that many researchers assumed that the HAB had benefits without operationally

defining the benefits being measured or the bond being examined. The results and discussions in

more recent publications continue to lack depth, direction, or replicability. These limitations may

be due to researcher biases, unclear theoretical rationales, or unreliable instrumentation.

Several potential mechanisms exist for the observed associations between pet ownership

and health benefits (McNicholas et al., 2005). First, pet ownership could exert a direct effect on

the owner’s health through the dynamics of the relationship between person and pet. This effect

has been compared to the positive physiological benefits of having interpersonal relationships

such as friendships on a person’s health (House, Lamberts, & Umberson, 1988). Another

explanation is that pets could enhance social contact and support, yielding an indirect effect on

the pet owner’s well-being. Another potential explanation could be that a third or outside factor,

34

such as personality characteristics or socioeconomic status could account for a person’s decision

to own a pet at all, explaining the observed connections between human health and pet

ownership. It may be also helpful to understand the process of competing-response theory

offered by Brickel (1982). Recent studies using physiological measures also have also provided

evidence of the HAB. Brickel (1982) suggested that the “generic ability of pet animals to

diminish anxiety and provide emotional support be theoretically subsumed within an extinction

model of learning, specifically that of competing-response theory” (p. 71). This theory informs numerous techniques in psychotherapy. In systematic desensitization or classic exposure, for example, a client is exposed to the aversive stimulus (or a variant of it) while simultaneously experiencing a stimulus that is neutral or pleasurable. The juxtaposed feelings cannot exist with equal intensity, the negative effects of the aversive stimulus are not brought to fruition, and the avoidant response can be extinguished with time and repeated exposure. Many classical

techniques include relaxation, mindfulness, or distraction techniques as the neutral or pleasurable

stimulus. It seems sensible that a pet could fill this role as well. Brickel notes, “Pets divert

attention from an anxiety-generating stimulus which the client faces. This interference allows for

self-monitored exposure to the stimulus instead of avoidance behaviors” (p. 72). Many

competing response stimuli have a single cue type (visual, auditory, or tactile, for example); a

pet, however, offers multiple cues, and may serve as an ideal stimulus for a competing response

paradigm in therapy. In many forms of therapy, cognitive distortions, problematic behaviors, or

disintegrated relationships are viewed as primary sources for an individual’s maladjustment.

According to cognitive theory, if thoughts or assumptions can be changed in favor of health,

automatic thoughts can be changed, schemas can be restructured, and a client’s overall level of

functioning will improve, as will life satisfaction (Stone, 1980). In many instances, individuals

35

find that avoidance behaviors are extremely effective, not only in not experiencing an aversive

stimulus but also in perpetuating and sometimes increasing the aversive nature of said stimulus

(See Figure 5).

As Brickel observed, pets also have the unique ability to create a deep psychosoical bond with their human counterpart. As many clients’ challenges are emotional in nature, that is, they avoid experiencing negative emotions, the presence of a client’s pet within the counseling office may offer a unique opportunity to move beyond distraction within exposure, and begin to address the more emotional aspects of the client’s experience, facilitating a longer, more sustained change over time. In the example of trauma exposure, for example, physical structures

of the brain are activated and continue to remain active when the client is “triggered” by a

memory of the event. In addition to using basic sensory distraction for exposure the presence of a

pet could touch an even deeper, more primal level of the brain, where the avoidance is really

taking place.

Virues-Ortega and Buela-Casal (2006) focused more specifically on health effects of pet

ownership in their meta-analysis of the human-animal interaction literature. They observed that

most published articles hypothesized three possible connections between pet ownership and

health: a direct correlation, an indirect connection through social facilitation, and outside factors

that simultaneously influence pet ownership and health indicators. In the area of direct effects,

they cited studies of cardiovascular response to interaction with a companion animal, and

clarified three primary hypotheses for the ensuing relaxation effect. First, they acknowledged

that tactile contact with another living organism could cause a reduction in cardiovascular

activity, such as a decrease in blood pressure. Second, they observed that unconditional

emotional support pets provide could have a buffering effect on the cardiovascular system.

36

Third, they cited classical conditioning and relaxation. The human participant could associate the companion animal with a healthy and calm atmosphere that would ultimately be associated with reduced cardiovascular activity. In their review of these studies, Virues-Ortega and Buela-Casal

(2006) noted that most published experiments were brief in nature and did not allow for

exploration of the long-term effect of pet ownership so they focused their meta-analysis on long-

term interaction with pets. Two patterns of research emerged: cross-sectional, between-subjects

designs contrasting pet owners with non-pet owners; and longitudinal, within-subjects designs

investigating the effect of pet on the owner. Unfortunately, the researchers were unable

to draw definitive conclusions about either type of study. The methods of each study varied

widely, as did participant demographics, physiological measures, and type of human-animal

interaction observed. Results were mixed and inconsistent, and neither author was able to

provide substantive reasoning for the discrepancies, again suggesting that a more precise and

strict research paradigm should be used to assess the effects of human-animal interaction (See

Figure 5).

In an extension of the classical conditioning hypothesis highlighted by Virues-Ortega and

Buela-Casal (2006), Gee, Church, and Altobelli (2010) asked preschoolers to perform

categorization tasks in the presence of a stuffed dog, a real dog, and a human to determine

whether any of the conditions led to enhanced performance on the cognitive task. Despite the

small sample size of 12 participants, the researchers found that preschoolers made significantly

fewer mistakes in the form of irrelevant choices when in the presence of the dog in comparison

to the human or stuffed animal, and concluded that the live dog’s presence could have had a

calming effect on the children, which enhanced their ability to make correct choices on the

37

cognitive task. The presence of the non-threatening animal enhanced student performance (Gee et al., 2010) (See Figure 5).

Again, numerous researchers (Davis, 1988; Imber-Black, 2009; Noonan, 2008) have extolled the beneficial roles of pets, citing their ability to reduce anxiety, depression, and overall stress levels, but these findings seem to garner little beyond a general sense that “pets are good.”

In 1984, Beck and Katcher conducted a critical review of the relevant literature on pet-facilitated therapy, and concluded that more scientifically rigorous studies were needed in order to more fully substantiate the claim that pets have a measurable, significant positive effect on humans.

They observed that, in many of these studies, investigators often have an inherent bias and will

“suspend critical judgment,” because they have a “feeling that pets are valuable and lovable” (p.

418). In addition to this internal bias, there often exists external pressure or anticipation that the research will be supportive of animal-facilitated interventions. The authors advocated a research design which would account for expectancy, novelty, objective measures, the use of controls, non-biased evaluation of data, and a cost-benefit analysis of the intervention (Beck, 1985; Beck

& Katcher, 1984). Other important suggestions (Johnson et al., 2002) stressed the importance of including both physiological and psychological measures in one’s study which are reliable and meaningful. They noted that a critical, yet elusive portion of the human-animal interaction and its accompanying research is dosage, which includes the “most beneficial timing, frequency, length, and content (p. 438)” of the specified intervention, particularly if non-significant results are observed, or if replication is to be possible. Much of the published literature suffers from lack of participant diversity, as the majority of respondents are Caucasian females who are able to manage their own finances. As Pachana and her colleagues observed, these demographic patterns could be meaningful and have the potential to play a role in the association between pet

38

ownership and well-being, but few researchers consider this in their analyses (Pachana et al.,

2005) (See Figure 5).

Yorke (2010) postulated that a strong and adaptive bond with an animal could “mimic the

impact of anti-depressant medication without the risks and side effects that might occur with

children who have experienced trauma” (p.566). Though this may seem to be a sweeping claim,

she traced pathways and neurotransmitter release in the limbic system, ventral pre-frontal cortex, orbitofrontal cortex, amygdala, and hippocampus. Yorke further identified processes by which touch, interaction, emotion and reward systems can lead to longstanding neural change for clients. Such change has the potential to encourage adaptive (healthy) neural development, enhance the brain’s resiliency to stress, or mitigate the neurological effects of trauma. Trauma also inhibits the flow of information between brain hemispheres, leading to an activation of the right brain and a deactivation of the left brain (Lipton & Fosha, 2011). Early relational trauma shapes the body, brain, and nervous system. Structurally, it may be seen as an overactive right amygdala or a shrunken hippocampus. Chemically, it may be manifested in high levels of cortisol. Behaviorally, it may be revealed as emotion dysregulation or inability to bond with and trust others securely (Lipton & Fosha, 2011).

39

Figure 5. Proposed Mechanisms of Action for Physiological Effects of HAB.

•Reduces blood pressure •Increased levels of oxytocin Direct Effect of Tactile Contact •Decreased levels of cortisol •Could act on neuroplasticity of brain, change pathways associated with touch

•Relaxation response, parasympathetic nervous system engaged Classical •Buffering effect on blood pressure Conditioning •Associate pet with social facilitation and support •Unconditional positive regard perception

Competing •Cannot experience relaxation and anxiety at same time, Response Theory so effect is masked

Figure 5. Three mechanisms which have been proposed for the physiological outcomes of

the connection between companion animals and human beings include direct effect, classical

conditioning, and competing response theory.

Effects on the companion animal. Some studies have focused on physiological effects

of human-animal interaction on the companion animal. Astrup, Gantt, and Stephens (1979)

measured dogs’ heart rates while being in a room with, and then petted by, different people and found that dogs responded differently based on their level of prior contact with the human participant. Human participants who had no prior contact with the dogs, indicating a lower level of bond to the animal, were associated with an increase in the dogs’ heart rates, which decreased over time and with repeated trials of petting. Animal technicians and individual pet owners had the strongest effect, with highest heart rates measured when in the same room but not in contact with the dog, and lowest heart rates while being petted, indicating that humans can have a calming effect when in physical contact with dogs, particularly animals they know well. Thus, it appears that the presence of a human, as well as tactile interaction, can have marked effects on a companion animal (Astrup, Gantt, & Stephens, 1979).

40

Odendaal and Lehmann (2000) found that both humans and dogs experienced beneficial

neurochemical changes in beta endorphins, beta phenylethylamine, prolactin, dopamine, and

oxytocin after 15 minutes of a positive interaction with one another.

Effects on humans- clinical samples. Other researchers focused on the physiological

responses of the human participants, and found statistically significant decreases in blood

pressure when participants touched a dog with which they reported feeling a bond (Baun et al.,

1984). These results were replicated in hospitalized children who interacted with dogs that had

completed training in pet therapy (Tsai, Friedmann, & Thomas, 2010). Similar effects were

observed when female research participants completed mental arithmetic tasks (Allen et al.,

1991) and when they read aloud in the presence of their own dogs but did not have physical

contact with them(Jenkins, 1986). These results indicate that companion animals have a

substantial impact on humans, just as dogs are affected by people.

A similar study of hospitalized heart failure patients noted decreases in blood pressure

(systolic pulmonary artery and pulmonary capillary wedge pressures), epinephrine, and

norepinephrine levels both during and after brief visits by therapy dogs, as compared to human- only visitation and control groups with no visitation (Cole et al., 2007). Additionally, participants who interacted with the dogs reported a significant reduction in anxiety, measured by the

Speilberger State-Trait Anxiety Inventory, when compared to human-only visitation and control

groups with no visitation. Demographic variables for the 76 participants were not analyzed, and

the authors did not conduct longitudinal replications, but the use of multiple groups indicated

that interaction with a dog had a significant impact on physiological and psychological

functioning in hospitalized patients (Cole et al., 2007). These findings have been supported in

longitudinal studies, with Friedmann and colleagues finding that one-year survival rates were

41 positively correlated with pet ownership among patients discharged from a coronary care unit irrespective of age (Friedmann, 1990; Friedmann et al., 1980; Friedmann & Thomas, 1995).

In 2007, Lust and colleagues measured changes in “as needed” medications in a rehabilitation facility following the introduction of a resident therapy dog. Fifty-eight residents participated in the study, ranging in age from 24-60 years and with various diagnoses, including traumatic brain injury, multiple sclerosis, and quadriplegia. Medication usage (for pain, constipation, and anxiety) and several physiological measures (blood pressure, body weight, respiration, and pulse) were tracked for three months leading up to the introduction of the dog and continued for nine months following his placement at the facility. Qualitative data from focus group interviews were also collected. Lust and her colleagues found that analgesic medication requests and pulse rates were significantly reduced over the course of the study. In addition, focus group interviews identified several quality of life improvements among the participants following the introduction of the dog, such as an increase in happiness and positive thoughts, an enhanced ability to think beyond one’s own needs, and feelings of pride, ownership, and normalization that the presence of the animal provided to the institution. The authors concluded that the introduction of the dog to a rehabilitation facility had a significant impact on patients in that pain medication requests decreased. Additionally, they observed pro-social behavior changes in their residents (Lust et al., 2007).

Effects on humans- non-clinical samples. Freidmann and colleagues (1983) applied this theory in a community sample of children, and hypothesized that the presence of a companion animal during various situations could have a significant physiological impact. More specifically, they predicted that the presence of a dog could reduce children’s’ blood pressure when reading aloud. Using a four-way ANOVA with repeated measures, they found that the dog

42

was associated with a significantly lower blood pressure, both in a condition of reading aloud,

and when the child was not performing any task, which supported Brickel’s (1982) theory.

Interestingly, the children did not speak to or touch the animals, and the dogs were unknown to

the children, yet the significant effect was observed. The animals’ presence, without additional

interaction, may have had a significant anxiolytic effect for the children.

A 2009 study by Miller and colleagues examined oxytocin level changes in pet owners as

a result of interacting with their dogs. The neurochemical oxytocin was selected due to its role in

bonding, stress relief, and socialization in human beings, and previous research indicating that an

individual’s oxytocin level increases following physical contact with dogs. Researchers

hypothesized that participants’ oxytocin levels would increase significantly following interaction

with their dogs, signifying an overall reduction in stress level and increase in affiliative behavior

between human and companion animal. Ten women and ten women and their pet dogs

participated in the study and serum oxytocin levels were measured in four conditions. In the

treatment condition, oxytocin levels were measured upon returning home from work and after 25

minutes of interaction with the participant’s dog. In the control condition, levels were measured

upon returning home from work and again after 25 minutes of reading nonfiction materials. All

participants were Caucasian, were employed full-time, and had owned the dog used in the study for a period of at least six months (Miller et al., 2009).

Results of the study were statistically significant and varied according to participant gender. Women in the study exhibited a 58.4% increase in oxytocin levels after pet interaction, and a 26% decrease in oxytocin levels after reading for 25 minutes. Men showed a 21.5% decrease in oxytocin levels after pet interaction, and a 56.3% decrease in the reading condition.

The authors attributed the gender differences to the physiology of the neuropeptide oxytocin and

43

its differential interaction with estrogen and testosterone, hormones whose levels vary

significantly according to gender, noting that estrogen exerts a stronger influence on oxytocin levels and may serve as a sex-specific biomarker of stress mediation in women only.

Additionally, the recruitment of a more diverse sample, as well as more than a single measure of stress, such as oxytocin and blood pressure, could have added depth and clarity to the results observed (Miller et al., 2009).

Shiloh, Sorek, and Terkel (2003) measured the effect of petting live and toy animals on anxiety levels in a stressful situation. In the laboratory experiment, 58 Israeli university students were put in a room with a tarantula spider and were informed that they might be asked to hold it.

Participants were randomly assigned to one of five groups: petting a live turtle, petting a live rabbit, petting a toy rabbit, petting a toy turtle, or petting nothing. Participants’ anxiety levels were measured by the Speilberger State-Trait Anxiety Inventory at baseline, after being informed about the tarantula, and after the petting condition, which lasted two minutes. The researchers also measured participants’ attitudes toward pet animals using Persky’s Companion Animal

Semantic Differential as a potential moderating variable. Among the participants, anxiety level was significantly higher than baseline after being told about the tarantula. In the experimental conditions, petting live animals yielded a significant reduction in anxiety scores compared to the control and toy animal conditions. The texture of the petted object (shell versus fur, for example) did not have a significant effect for either the live or toy condition. Additionally, attitudes toward pet animals did not have a significant moderating effect on anxiety level, suggesting that, regardless of one’s attitude or affiliation toward animals, the presence of and physical contact with a live animal may reduce one’s anxiety level in a significant and measurable way. No data were collected on longer term effects of interaction, and the study was limited to a restricted

44

range of ages and ethnic backgrounds. The results offered evidence for the anxiolytic impact of

interacting with a living animal during a stressful situation (Shiloh, Sorek, & Terkel, 2003).

In an effort to make sense of the physiological data and studies involving human-animal

interaction (HAI), Beetz and colleagues (2012) completed a literature review of 69 original

studies which examined the effects of interaction between humans and animals. All studies in

included had a minimum of ten participants and documented that the observed effects were due

to HAI and other confounding variables. The studies included the following variables: social

interaction, social stimulation, trust, empathy, aggression positive affect, depression, anxiety,

fear, pain, cortisol, epinephrine, norepinephrine, blood pressure, cardiovascular health, and

learning. In their review, the authors found evidence for a potential link between oxytocin and

HAI, and proposed that many of the published effects across studies could be at least partially

explained by the operation of the oxytocin system, rendering it a core mechanism in

understanding the HAB. The authors did not perform a meta-analysis of the data, and no overall effect size was reported (Beetz et al., 2012).

Evidence of the Psychosocial Benefits of the HAB

Just as some researchers have investigated the physiological effects of the HAB, others have focused more exclusively on the psychosocial aspects of the connection between person and pet. Some (Speck, 1964) hoped to understand a pet's symbolic role within a family, while others measured the outcome associated with introducing a companion animal to a novel environment (Davis, 1988; Esteves & Stokes, 2008; Sockalingham et al., 2008). Still others worked to understand the extent to which pet ownership could be associated with an individual's personality or behaviors (Bagley & Gonsman, 2005; Johnson & Rule, 1991; Knight & Edwards,

2008).

45

In a 1964 article published the American Veterinary Medical Association Journal, Speck appealed to the veterinarian’s capacity “to understand certain aspects of his client’s family life through his understanding of their pet” (p. 150). Speck, a psychiatrist and family therapist, identified numerous links between family members and their companion animals. His theoretical foundation was psychodynamic and symbolic in nature, and noted that pets often serve as barometers of family functioning. For example in “disturbed” , Speck observed that pets sometimes die, “As if a ‘sacrifice’ has occurred” (p. 152). As such, an examination of the physical health and general functioning of the pet could serve an important role in exploring the health of the client and his or her home life. Bringing the client’s pet into the counseling session, whether physically or metaphorically, could provide yet another important piece of information about the client’s world. Such discussions could provide a helpful context for the therapist, illustrating interactions between family members, discussions about the pet and general treatment of the pet from the client’s or family members’ points of view.

Yorke (2010) contended that attuned connections between human beings can be classified as social networks, “Unique to each individual enhancing their feelings of competency, self-worth and self-efficacy” (p. 563). She further shared that these connections can serve as buffering agents, fostering resiliency, healthy bonding, and normal neural development throughout the lifespan. Citing decades of research on the HAB, Yorke traced evidence that the bond between humans and animals is neurophysiological in nature (See Figure 6). She further concluded that human-animal relationships rely on attachment and affiliation, and “appear to have some impact on neurochemical processes in the brain” (p. 564). In psychotherapy, recent advances in neurophysiological research suggest that counseling interventions, particularly in the area of trauma, can be traced to structural changes in the brain. Both pet ownership and

46

behavioral psychotherapy (including meditation) can lead to physical, long-lasting changes to the

brain (e.g., Chinner, 1991; Sockalingham et al., 2008; Virues-Ortega & Buela-Casal, 2006;

Walsh, 2009).

Levinson was a psychodynamic therapist who utilized his pet dog Jingles in counseling sessions. He observed his clients’ behaviors toward Jingles and interpreted their meanings. As he acknowledged, however, his observations were serendipitous in nature, rather than being grounded in formal methodology [among his interpretations: “The cat is preferred by some children who are undergoing oedipal conflict and are resentful of authority.” (p.245)] He published numerous manuscripts on the subject, and believed deeply in the healing power of bonding with a companion animal, but he did not utilize a systematic process for reviewing and understanding his observations. Levinson advocated for the development of more rigorously planned techniques for measuring the phenomena he witnessed, but did not pioneer or outline such approaches (1965). He noted that companion animals provided clients with unconditional positive regard and often facilitated communication between the client and therapist. This could mean that the presence of a pet, whether in session or in the home life of the client, could have the potential to provide meaningful and adaptive support and attachment (See Figure 6).

Companion animals are omnipresent sources of physical, social, and emotional support.

As Levinson (1964) noted, the ownership of a pet comes with a certain level of responsibility and emotional maturity on the part of the human owner. Beyond providing the basic necessities of food, water, and shelter, the owner often has expectations for pleasant interaction.

Additionally, a certain level of affiliation arises when two or more living being coexist. The pet is often named, obedience or housetraining is usually implemented, and systems of play or recreation are employed. For many individuals living with mental illness or experiencing

47

significant life distress, a pet can serve as a stabilizing presence, reducing social isolation and

providing reciprocal affiliation, as defined by Davis (1988). In addition to this, pets offer “love

and tactile reassurance without criticism and… perpetual infantile innocent dependence that may

stimulate our natural tendency to offer support and protection” (Corson et al., 1975).

The nursing profession has long recognized the potential role of pets in healing. More than 150 years ago, Florence Nightingale offered an alternative prescription, noting that “A small pet is often an excellent companion for the sick or long chronic cases, especially” (Nightingale,

1859, p. 58). Davis (1988) recounted several nursing surveys among long-term inpatient facilities and retirement homes. In addition to considering pets to be family members, these patients identified pets as the “most missed” possessions, ranking them above and other valuables. Pets often take on an anthropomorphic role, and “are believed to embody love and acceptance, promoting a sense of reciprocal affiliation” (Davis, 1988, p. 79) (See Figure 6).

Figure 6. Proposed Mechanisms for Psychosocial Benefits of HAB.

• Pet plays symbolic role in family functioning and Psychodynamic • Pet as object for projection. protection

• Associate pet with social facilitation and support Social Facilitation • Unconditional positive regard perception, which then enhances well-being

Figure 6. Two primary mechanisms proposed for the psychosocial benefits of the connection between humans and animals include psychodynamic theory and social facilitation.

Effects on humans- clinical samples. Some studies on the HAB provide short synopses of observations but assert wide-reaching claims of health benefits without critically analyzing the literature (Ormerod et al., 2005; Ovesen, 2005). Davis (1988) offered a brief description of

48

animal-facilitated therapy in nursing, but relied heavily on descriptions of having pets visit

facilities or case studies to justify the implementation of pet therapy programs in nursing

facilities. Socklingham and colleagues (2008) offered an in-depth case study of a 43-year-old single male diagnosed with bipolar disorder who had experienced a physical assault and presented to treatment with low self-confidence, depressed mood, and lack of independence. A therapy dog was added to his treatment regimen, and the researchers observed a marked shift in his overall well-being, but failed to provide evidence of the changes in the form of behavioral analyses or changes in symptomatology according to validated measures of well-being. Zimolag and Krupa (2010) reported a similar account, in which a client with bipolar disorder credited her relationship with her companion animal as critical in helping her move beyond stigma by providing continuity, belonging, acceptance, and a means for community integration. This study applied a theory of community integration and stigma reduction, but again did not utilize standardized scales or measures to quantify changes over time (Sockalingham et al., 2008).

A 2004 study of seven individuals with schizophrenia living in a Hungarian psychiatric

hospital offered additional support for human-animal interactions and interventions. Over a nine-

month period the individuals attended weekly therapy with a dog present, and independent living

skills were assessed before and after the intervention period, with the finding that health and

domestic activity social skills improved significantly. Grooming, eating, transportation, money

management, and leisure social skills also improved, but none of these improvements were

statistically significant, suggesting that the introduction of pets may have an impact with an

inpatient population but that the instrumentation may not accurately reflect the nuanced and

individualized nature of the connection between humans and animals (Kovacs et al., 2004).

Sellers (2006) conducted a similar study with four long-term inpatient individuals with dementia.

49

The intervention consisted of five, 15-minute interactions with a dog, and measurements included social behaviors (using the Social Behavior Observation Checklist) and agitated behaviors (using the Agitated Behavior Mapping Instrument). Sellers used a repeated experimental design, and observed reductions in agitated behaviors and increased pro-social behaviors that were statistically significant. Both studies had similar limitations: small sample size hindered the statistical power, as did the lack of multidimensional, validated measures of well-being or other methods of assessing behavioral change in the participants over the course of the intervention or beyond the treatment period. It is unknown whether these changes were maintained over time (Kovacs et al., 2004; Sellers, 2006).

In situations where a pet owner is exposed to trauma, the companion animal can take on a vital role. Flynn (2000) hypothesized that women who experience and own a pet could offer a unique perspective as to the role and function of the pet in the home. Surveying

107 clients at a battered women’s shelter in South Carolina, he found that nearly half of the respondents (46.5%) reported that their companion animals had been threatened or harmed by their partners. Additionally, pets were viewed as providing emotional support to 75% of all the women, and 90% of those who indicated their pets had been abused. One out of five women in the study noted that she delayed seeking shelter because she was concerned about the physical and emotional welfare of her pet(s). In follow-up qualitative interviews with ten of the respondents, clear themes were identified: pets were considered to be family members, inflicted on pets had myriad emotional consequences for the women and children who bore witness, and attributions of power and control were among the reasons women believed their pets were abused. In examining the animals’ responses to the women’s traumas, the interviewees noted that their pets served as comforters following each assault, seeming to listen when the

50

women confided in them and providing physical warmth and affection in the aftermath.

According to the women, pets also took on the role of protector, positioning themselves between

perpetrator and victim or providing some form of distraction in times of tension. Pets and their

owners appear to share a deep bond, particularly in the face of trauma. In a chaotic family

situation where trust is uncertain and danger is omnipresent, this connection may enhance one’s

ability to feel grounded and safe (Flynn, 2000).

Esteves and Stokes (2008) used an experimental repeated measures design to measure the

psychosocial effect of introducing a dog to a special education classroom, and noted that the

presence of the dog led to an increase in positive verbal and nonverbal behaviors among the

students, a decrease in negative social behavior, and a generalization of improved social

interaction among the students. The dog had not yet completed therapy dog training, and the

sample size was three students, which limited the ability of the researchers to perform

quantitative analyses of observed behaviors or draw meaningful conclusions about the

mechanism of action for the dog’s presence or efficacy. Again, the decision to use observation

instead of standardized measures made it difficult to generalize the findings or begin to

understand the role and mechanism of the HAB (Esteves & Stokes, 2008).

In therapy, an important issue is attendance and participation on the part of the client.

Beck and colleagues (1986) found that the presence of finches in psychiatric inpatient group

therapy was associated with higher levels of attendance, more vocal participation in the sessions,

and exhibited fewer hostile behaviors in comparison to a control group, which met in the same

room with the same facilitators. The study concluded early, in fact, because one-half of the experimental group (which featured birds present during therapy) had left the group due to symptom amelioration. This was one of the first rigorously controlled experimental studies of its

51 kind, and it featured clear and consistent results. The birds did not interact with the participants but instead were caged and placed in one corner of the room. (Beck et al., 1986).

In 1990, Draper, Gerber, and Layng attempted to define the role and effect of animals in individual psychotherapy using an exploratory approach. Clients were between the ages of 20 and 87 years, and had various psychiatric diagnoses, including mental retardation, dementia, and traumatic brain injury. All had been previously described as “withdrawn and demonstrated little to no spontaneous speech” (p. 170). The 10 clients were asked in the therapy session if they would like to see and interact with a dog which was in a screened cage in the counseling room.

All participants agreed, and were allowed to interact with the animal. Three sessions were videotaped and reviewed for each participant, and raters observed clients’ affect, communication, and behavior at three points in each taped session: prior to the introduction of the dog, after the animal’s introduction, and at the midpoint of the therapy session. The researchers observed the following within and across sessions: progressive increase in the duration of attention to therapist and progressive decrease in attention to animal. The findings were not statistically significant, but the authors noted that the trend was encouraging, with the concept that the “animal…acted as a prosthesis facilitating the establishment of a therapeutic relationship by the therapist” (p. 172).

The authors acknowledged that the existence of a vetted, reliable measure would have been preferable to the checklist they created, and they reported a low level of interrater reliability, which could have negatively impacted the results. Additionally, two minutes of each condition were observed, as opposed to the entirety of the tape, so the observational data was not as rich as it could have been.

In a comparison of inpatient substance abuse group therapy with and without an animal- assisted therapeutic intervention, Wesley, Minatrea, and Watson (2009) found that the presence

52 of an animal was associated with a more positive opinion of the therapeutic alliance, as measured by the Helping Alliance Questionnaire. Half of the 231 participants were female, 90% were

Caucasian, and more than 50% were under the age of 25 years. Participants were randomly assigned to the experimental (with therapy dog present) or the control condition, and participated in 14 sessions of group therapy. The researchers concluded that having the pet present during the intervention was positively correlated with a greater trust and belief in counseling effectiveness, which is of clinical importance because greater trust in a counselor has been associated with higher levels of sobriety in individuals with chemical dependency issues (Wesley et al., 2009).

Other researchers have acknowledged that many therapists use their own pets in counseling, and sought to understand the reasons and effects of such a practice. In a qualitative, exploratory interview with 13 practicing therapists, Mason and Hagan (1999) found that many therapists’ initial motivation for bringing their pets to the office were personal in nature, as opposed to being purposeful or client-guided. A few of the counselors had read anecdotes about the role that animals could play in therapy sessions, but most disclosed that they wanted to have their companion animals with them in the office because they enjoyed the company the pet provided. The therapists stated they most often used their pets in sessions with clients who were female, White, or children, but noted that they did not consider the diagnosis as heavily in the inclusion criteria, so pets were used with clients having various disorders. More than one-third of the interviewed therapists stated that some clients feared or disliked dogs, and this became a focus in therapy. Most often, processing the fear and anxiety yielded a significant feeling of achievement for these clients, according to therapists (Mason & Hagan, 1999). As early as 1965,

Woods noted the importance of clients’ pets in conceptualizing and understanding one’s life

53

experiences; however, systematic efforts to bring pets into therapy, whether physically or

symbolically, has not taken place.

Effects on humans- non-clinical samples. Other theoretical constructs cite the use of

touch, nurturance, and interpersonal interaction to establish healthy adaptation in working with

trauma survivors; “Soothing, holding, rocking, creating safety are examples of behaviors that

move children to this [safe] place” (Yorke, 2010, p. 565). Animal-assisted interventions lend themselves well to this area. Equine-assisted interventions (EAIs) provide the structured interaction and safety in ways that foster trust, safety, and emotional attunement between horse and rider, by providing body to body contact with a large animal, and taking place in a novel, enriching environment. EAIs have had the following outcomes: increasing self-esteem, enhancing one’s sense of personal power, and reducing antisocial behavior (Yorke, 2010).

Yorke, Adams, and Coady (2008) also added depth to the concept of the HAB, but focused more specifically on the bond which could arise from interaction between trauma survivors and horses.

Their exploratory qualitative study involved naturalistic observation of human-equine interaction, as well as semi-structured interviews with six individuals who had trauma backgrounds (four females and two males). The participants were selected because of their previous experience with horses and their decision to ride horses following their traumatic exposures. The participants were interviewed about both their trauma and about the perceived impact of the human-horse relationship in facilitating recovery. Yorke, Adams, and Coady also videotaped the interactions and interviews, and then participants’ visual behaviors and words were coded and analyzed. The results of the data analysis yielded primary themes which the researchers fit to a theory of healing. They theorized that the bond was comprised of both a task dimension and an emotional dimension. The emotional dimension included an intimacy bond,

54

wherein the person believed he or she experienced nonjudgmental love by the horse, and an

identity bond, wherein the person could readily identify his or her role in the relationship with

the horse. The task dimension included a partnership bond, as defined by the person’s belief that

he or she was in equal partnership with the horse, and a utility bond, acknowledging that

meaningful activities were completed as a result of the human-horse partnership. All four

components could then have a significant role in healing. A strength of this study was the

authors’ focus on already-existing interactions, as opposed to asking participants to enroll in a program in which bonds would need to be created. This decision, and the authors’ accompanying discussion, may be more similar to the pet-owner relationship than pet therapy or equine-assisted therapy. Again, while this work could provide a helpful frame of reference or beginning point for understanding and quantifying the HAB in the future, Yorke, Adams, and Coady (2008) did not use standardized measures, and they did not provide insight about how to operationalize the four components of partnership bond, utility bond, intimacy bond, or identity bond in future studies.

Knight and Edwards’ (2008) qualitative study of dog owners’ perceptions of companion animal benefits did offer additional insight about the HAB. The researchers recruited 62 dog owners from popular dog-walking locations in the United Kingdom to participate in focus group discussions pertaining to beliefs and attitudes about pet ownership and human-animal interaction.

Participants were primarily female (76%), retired (average age was 60 years), and had owned dogs for most of their lives (84%). Additionally, 97% of the focus group participants reported walking their dogs at least once per day, and cited this physical activity and the provision of safety and security as primary benefits of dog ownership. Other key benefits of dog ownership included psychological rewards such as receiving , decreased loneliness, and the idea that dogs could serve as “therapists” and “family members,” providing comfort and

55 motivation to move on in times of sadness and loss. Knight and Edwards also acknowledged the interrelated nature of these physical, psychological, and social benefits, noting that it was often difficult to separate these effects in data analysis and suggesting that typical quantitative assessment instruments may not fully capture the multifaceted experience of dog ownership.

Despite this, the benefits listed by the participants could provide a beginning point for additional studies of the HAB (Knight & Edwards, 2008).

Wiggett-Barnard and Steel (2008) focused more specifically on the experience of guide dog ownership with six individuals from South Africa in their qualitative study. Using a structured interview with five men and one woman, all of whom were legally blind, eight themes emerged about owning a guide dog: an improvement in mobility, the provision of companionship, an increase in independence and personal agency, a stabilization of lifestyle and financial choices, an increase in social interaction, an awareness of the dogs’ imperfections and levels of distractibility, an awareness that the public lacks education about guide dogs, and a deep appreciation and pride in owning a guide dog.

Antonacopoulos and Pychyl (2010) sought to understand the role of anthropomorphism

(the degree to which a person can assign human characteristics and traits to an animal) and social support as mediators of physical and psychological health in dog owners. They hypothesized that owners with low levels of human social support would have poorer health if they engaged in high rather than low levels of anthropomorphism, while owners with high levels of human social support would not vary depending on their anthropomorphism levels. The researchers tested these hypotheses using an online survey and of 203 Canadian dog owners. The sample was similar to other published studies involving the HAB, in that 90% of participants were female, which limits the generalizability of the results. Racial and ethnic background information was

56

not collected, but respondents ranged in age from 18 to 62, and were varied in the following

areas: marital status, education level, and income level. Measures included: Albert and Bulcroft’s

10-item anthropomorphism scale, Zimet, Dahlem, Zimet, and Farley’s 12-item Multidimensional

Scale of Perceived Social Support (MSPSS), Cohen, Kamarack, and Mermelsterin’s 14-item

Perceived Stress Scale (PSS), and Radloff’s 20-item Center for Epidemiological Studies-

Depression Scale (CES-D). In addition, the participants reported the number of medical visits,

prescription medications, and self-reported level of physical health over the past year. They also

indicated how much time they spent engaged in physical activity and spent outdoors in the past

week, as well as how many years they owned their oldest dog. Results indicated that dog owners

with high levels of human social support did not differ significantly in their number of doctor

visits or number of medications regardless of whether they engaged in low or high levels of

anthropomorphism. Among owners with low levels of human social support, those who engaged

in high levels of anthropomorphism had significantly more doctor visits and took significantly

more medications than those who engaged in low levels of anthropomorphism. Health ratings did

not differ significantly according to whether dog guardians engaged in low or high levels of

anthropomorphism or social support. The study was limited in that no control group of non-

guardians was surveyed. As a result, causal relationships could not be inferred or tested.

Additionally, the sample was homogeneous in nature with regards to gender, as are many of the

published HAB studies (Antonacopoulos & Pychyl, 2010).

Adamle, Riley, and Carlson (2009) gauged college students’ interest in participating in a

pet therapy program. Participants in the study self-selected into workshops about pet therapy as part of a university’s orientation program for first year students. Students completed questionnaires about their past pet ownership and interaction with animals, then learned about

57 pet therapy as a complementary or alternative support method. After the presentation, six dog- human handler pairs entered the presentation, answered questions, and interacted with the students. The 246-participant sample was 98% single, 91% Caucasian, and 85% female.

Additionally, 91% had pets at home, 92.5% considered pets to be an integral part of life, and

90.3% reported that pets provided comfort in times of stress. After taking part in the workshop,

96% of participants indicated interest in a pet therapy program on campus. Though the results were significant, the study had numerous limitations, given the homogeneity of the participants, their self-selection into the study, and the researchers’ use of a “yes” or “no” questionnaire instead of a Likert-type assessment. No control group was used, but the addition of a workshop without the dog-human handlers or using the survey with a different workshop (such as women’s health issues) could have provided a more heterogeneous sample that could allow broader conclusions. Additionally, clearer measures assessing the HAB could have added depth to the study (Adamle, Riley, & Carlson, 2009).

In a study of 194 undergraduate students, Taylor and Signal (2005) found a significant relationship between human empathy and attitudes toward animals, regardless of gender, though females scored higher on positive attitudes toward animals. Marks, Koepke, and Bradley (1994) surveyed 179 university students and found that the HAB (as measured by the Pet Attachment

Survey for pet owners and the Pet Attitude Inventory for non-owners) was positively correlated to generativity, conceptually defined as concern for future generations (measured by the Loyola

Generativity Scale).

Brown and Katcher (2001) focused more specifically on the HAB and dissociation, with three purposes: replicating their previous findings of a positive correlation between companion animal bond and dissociation in a sample of veterinary technician and college students;

58 replicating their previous research that individuals with high levels of pet bond are three times more likely to experience clinical levels of dissociation than those with low pet bond; and to compare pet bond and dissociation ratings of veterinary technician students to other samples of participants. Participants were 120 American females enrolled in a two-year veterinary technician program. All subjects were over 18 years old, 95% were Caucasian, and 98% currently owned a companion animal. Participants completed the 28-item Dissociative

Experiences Scale (DES) and the Pet Attachment Questionnaire (PAQ). They found that a moderate and consistent positive correlation between pet bond and dissociation. Additionally, participants with higher pet bond scores were much more likely to have clinical levels of dissociation (43%) than those subjects with lower pet bond (0%), suggesting the frequency of serious dissociative symptoms was significantly higher in people who were highly attached to their pets (Brown & Katcher, 2001).

Brown and Katcher (2001) postulated that high dissociability could predispose someone to becoming strongly bonded to pets. A possible explanation is that high levels of dissociation usually correlate with childhood abuse, sometimes manifested later in life as distrust of humans.

Relationships with pet could replace interpersonal relationships or provide a bridge to begin relating to others. Another explanation for the study’s results could be imaginative involvement or fantasy proneness, as defined by the Dissociative Experiences Scale, which captures the enjoyment of participating in adventurous or fantasy activities similar to hypnosis or relaxation exercises (Brown & Katcher, 2001).

Cline (2010) focused specifically on the relationship between dog ownership and well- being. Cline hypothesized that dog ownership could decrease depression levels by providing social support and physical activity. She further hypothesized that moderators of the relationship

59

were marital status, gender, and age. Using random digital dialing and a 45-60 minute, 200-

question interview, 201 adults between the ages of 19 and 94 years were recruited for

participation. The following demographics were observed in the national sample: 86% were

Caucasian, 70% were female, 61% were married, and 38% owned at least one dog. Cline used

the Center for Epidemiologic Studies Depression Scale (CES-D) to measure depression and negative well-being. Social support was measured with a single question by asking participants to rate on a scale of one (not at all satisfied) to four (very satisfied) how they felt about the emotional support they received from friends and family. Physical activity was also reported in a single question ranking from zero (never or once per week) to five (more than once per day).

Using an ordinary least squares regression, Cline found no evidence that dog ownership and well-being are mediated by age, physical activity, or social support. She did find evidence that

dog ownership was associated with greater well-being for women and lower well-being for men, and that the dog ownership was associated with greater well-being for single individuals and lower well-being for married individuals. The sample was not nationally representative in terms of age, gender, or race, which limits the generalizability of the findings. Additionally, Cline used single questions to measure constructs such as physical activity, relationship status, social support, and pet ownership. Cline’s results could have been more descriptive and meaningful if reliable measures of these constructs had been used to learn more about current and past pet ownership, the HAB, and the participants’ physical health (Cline, 2010).

Cohen (2002) sought to understand the role of pets in urban American families, as well as to explore how that role compared to that of human family members. Her research was conducted in two phases. In the first phase, 201 randomly selected pet owners at a veterinary teaching hospital in New York City were asked identical questions about the human family

60

member and the pet to whom they felt the closest in a 15-minute survey. In the second phase, a

subgroup of 16 participants completed a social network instrument and answered questions

designed to compare feelings about pets and human family members to whom the participants

felt close in a two-hour interview.

In the first phase of Cohen’s study, 75% of participants were female, 66% were college graduates, and 54% were married or living with a partner. Participants’ ages ranged from 18-85, and 35% percent had at least one child. More people had dogs (55%) than cats (27%), though some had both (18%). Seventy-three percent of respondents identified themselves as non-Latino

Caucasian, 3% as African American, 5% as Asian, and 12% as Latino; 7% did not identify themselves as belonging to any ethnic group. The researcher compiled a questionnaire from existing, validated instruments to assess four primary areas in phase one: so participants could compare their relationship with the human family member and the pet to whom they felt closest, to allow participants with children to compare those two relationships, to gauge whether people with close pet relationships are lonely or dislike other people, and to measure the strength of the bond between participants and pets. Cohen used the 11-item family subscale of the Revised

Kinship Scale, 16 items from the 17-item Intimacy Scale, four questions from the 20-item UCLA

Loneliness Scale (RULS), four questions from the Social Fear Scale (SFS), six questions from the 25-item Index of Parental Attitude (IPA), and the eight-item Companion Animal Bonding

Scale (CABS). Reliability for each scale was pilot tested on a similar group of adults and then computed using all 201 questionnaires. The most significant single predictor of the score for any scale was the number of hours spent with pet and the subsequent score on the Poresky

Companion Animal Bonding Scale (p = .0001). Cohen found that participants who spent 16 or more hours a day with their cat or dog scored significantly higher in bond, intimacy with pet, and

61 with pet than the rest of the participants. She also found that spending most of the day with a pet was not related to loneliness, but it was significantly correlated with higher levels of social fear. Women expressed more positive feelings about all of their relationships than men.

Women reported feeling significantly more intimacy and kinship with pets than the male participants. The women also reported fewer problems with pets and less overall loneliness than men did. Being a college graduate was inversely related to feelings about relationships. College graduates scored significantly lower than non-graduates on psychological kinship and intimacy with both people and pets (Cohen, 2002).

Gender and education level had the greatest influences on answers to questions in the first phase, so in the second phase of 16 participants, Cohen selected four male college graduates, four male non-graduates, four female college graduates, and four female non-graduates.

Additionally, she ensured that at least one participant with a cat was in each subgroup of four people. Beyond these restrictions, the demographic variables from the first phase were repeated in the second phase. In the second phase, Cohen used The Social Network Map and Grid

(SNMG), then asked respondents questions to learn more about the participants’ sources of support in a semi-structured interview. Eighty-two percent (13) of the participants in the second phase of the study reported their pets were major sources of emotional and social support, and an equal percentage acknowledged that they relied heavily on their human family circle for concrete, informational, and emotional support, with no significant differences attributable to gender or educational level. Cohen concluded that companion animals serve as family members, but are viewed as separate from the family. She noted that her assessment instruments, while validated, were adapted for use with pets, which could have impacted her results, as she made an assumption that pet bond could be considered equivalent to human attachment style. The study’s

62

sample was primarily female, urban, and educated, which hinders external validity and

generalizability (Cohen, 2002).

In an effort to focus more on participant diversity, Brown (2003) assessed differences

between Caucasian and African American veterinary students in terms of degree or strength of

the HAB, as measured by the Pet Attachment Questionnaire (PAQ). She found that Caucasian

students were more likely to own pets, owned a greater number of pets and greater variety of

pets. They were also more likely to allow pet to sleep on bed, and reported higher overall scores of what they identified as attachment to pet, which signified the strength of HAB. Again, the theoretical argument for equating attachment theory to the HAB was not built by the author; as a result, the use of the phrase “pet attachment” may be a misnomer (Brown, 2003).

Figure 7. Summary of HAB Benefits.

Physiological Benefits Psychosocial Benefits

•Decreased Anxiety in clinical and non-clincal •Increased social skills samples •Increased attendance and participation in •Decreased Blood Pressure in clincial and non- therapy clincal samples •Enhanced perception of therapeutic alliance •Increased release of Oxytocin in females in •Increased self-esteem and independence non-clincial samples •Increased perceived social support •Enhanced overall well-being •Decreased loneliness

Figure 7. This figure provides an overall summary of empirically researched benefits of the bond

between humans and companion animals, from both physiological and psychosocial foci.

A Crossroads: The HAB and Attachment Theory

Construct Equivalence?

Multiple authors have used the terms HAB and attachment interchangeably, assuming

they are parallel concepts. As noted by Kobak (2009), “researchers have not entirely settled on

63 criteria for defining attachment bonds across the lifespan, nor have they developed a systematic validation of measures for assessing relationships as attachment bonds" (p. 447). Geist (2011) for example, used applications of human attachment and cognitive theories without adequately building an argument that these concepts were equivalent or meaningful when applied to animals. Other authors provide anecdotal reports of people’s attachment to pets, without operationally or theoretically defining the construct of pet attachment (Antonacopoulos &

Pychyl, 2010; Berryman et al., 1985; Brown, 2003). In what may have been the earliest published attempt to explain the HAB in terms of attachment theory, Rynearson (1978) provided case examples of clients’ interactions with and beliefs about their pets, then categorized the dyads in terms of attachment style without any formal measurement. Sacks (2008) used a similar format when presenting her rationale for bringing her own companion animal into psychoanalytic therapy sessions with clients. She did not, however, consider the therapeutic value of allowing clients to bring their own pets into the environment which, as will become evident, may provide a better illustration of traditional attachment between pets and humans, given the novelty and possible threat of the experience for the client.

Crawford, Worsham, and Swinehart (2006) posited that the following factors should be assessed systematically in exploring the HAB prior to assuming that attachment to pets and humans are parallel concepts: emotional bond, goodness of fit (how the dyad functions together), secure base, seeking proximity, and representational models (such as organizational rules about experiences related to attachment, such coping with stressful events). Shore, Douglas, and Riley

(2005) noted that low attachment scores on published measures such as the Lexington

Attachment to Pets Scale (LAPS) do not necessarily mean that the owner does not have positive feelings toward or provide basic care (food, medical attention, shelter) for their pets.

64

As the human-animal interaction (HAI) literature has matured, some researchers have

begun to distinguish between the HAB and the concept of attachment between owners and

companion animals. Crawford, Worsham, and Swinehart (2006) noted that one must distinguish

“between attachment as measured by current research on the human-companion animal relationship and as defined…by attachment theorists using… standard methods of assessing attachment” (p. 99). They hold that current use of the term in attachment literature refers to an individual “seeking or maintaining proximity to another, clearly specified individual, who functions as a secure base and who is perceived as better able to cope with life stressors” (p, 99), whereas human-animal attachment describes a relationship which is hierarchical and emotional in nature. They further suggest that the purest way to apply formal attachment theory to the HAB would be through the Strange Situation and the Adult Attachment Interview. The ASSP and AAI are qualitative in nature, whereas many of the HAB studies have been quantitative. Published scales such as the Companion Animal Bonding Scale (CABS), the Pet Attachment Survey (PAS) and the Lexington Attachment to Pets Scale (LAPS), focus on the owners’ feelings toward pets, rather than observation of behavior in situations showcasing attachment style. These scales do not provide a thorough assessment of an individual’s attachment to a companion animal.

Pets as Attachment Figures

In pet ownership, a meaningful bond between pet and owner has been compared to that of parent and child, and could explain the significant physiological effects that a pet's presence has on the owner's functioning. “The therapist’s role is to facilitate the establishment of a secure base with the client, which makes it possible to investigate and re-experience various aspects of one’s life, some of which would be too painful to think about without a reliable and trusted companion” (Sable, 2004, p. 11).

65

Attachment Behaviors. Sable (2004) described the process of disrupted attachment

through an informal exploration of two pet dogs she adopted from the same litter. She detailed

the behaviors and responses of one of the dogs in particular, noting that she stiffened and had

blank eyes when touched for the first time, and immediately hid as soon as she was no longer

being touched. According to Sable, the dog displayed an avoidant attachment style and asserted

that such a style could be overcome within a trusting, safe relationship. The climate she

described in training her pets was remarkably similar to Rogers' approach to psychotherapy, in

that it was warm, genuine, empathic, and comforting, while allowing the animal to explore his or

her own needs and establish more effective working models of the world around him or

her. Sable traced her interactions with the dog over the course of their relationship, and she

observed that the dog’s connection to her seemed to strengthen and become more secure over

time. The evidence Sable cited suggests that attachment could be a useful framework for

conceptualizing human relationships with pets, but she did not systematically apply the theory to

measures or studies (Sable, 2004).

Jasperson (2010) highlighted the similarities of the attachment behaviors of dogs and human infants, noting that they have the same needs for physical, emotional, and psychological sustenance and nurturing. She piloted an animal assisted group therapy program in a women’s correctional facility. Over the course of eight, one-hour sessions, the group of female inmates covered the topics of boundaries, personal safety, trust, being trustworthy, responsibility, understanding emotion, expressing emotions appropriately, and learning new behaviors for daily functioning. All five participants had psychiatric diagnoses comorbid with substance abuse or dependence. A certified therapy dog was present for each session, and served as the model or teaching tool for each topic. Again, the dog provided a corrective emotional experience for the

66

women, serving as a tangible model for practicing effective interpersonal communication and

other psychoeducational topics (Jasperson, 2010).

Noonan (2009) offered the description of pet as transitional object, serving as mediation

tools for the pet owner and the outside world. She noted that pets are “one maturational step

beyond the transitional object” (p.400), serving as a mirror of family interplay and dynamics.

They may also serve as objects upon which one may project. Another study involving children’s

interactions to robotic dogs and living dogs illustrated these conclusions (Melson et al., 2009).

Researchers observed children aged 7-15 years interacting with AIBO, Sony Corporation’s robotic dog and compared this to interactions with a living dog. Children treated the robotic dog as an artifact worthy of exploration, and offered the living dog significantly more affection in terms of touching and petting.

Attachment Bonds. In describing animal companions, Margolies (1999) noted, “Pets are

devoted, forgiving, affectionate, uncritical, and available. Their love is given unconditionally”

(p. 298). Therapists such as Carl Rogers (1980) have used similar language to characterize the

role of the psychotherapist as present, accessible, congruent, accepting, transparent, and empathetic. Levinson (1965) was one of the first psychotherapists to publish on the use of pets in therapy. He brought his own companion animal, a dog named Jingles, to work with him, and noticed profound changes in his patients’ reactions. Often the children became more engaged and verbalized more when the dog was in session. Levinson reasoned, “If we look these [behavior] disorders as maladaptive response patterns, then using pets to help modify these patterns and aid children in reorganizing their ‘style of life’ appears reasonable” (p. 695). Using a psychodynamic lens, he theorized that the bond a child has with a pet has many effects, noting that a client could

“conceive of a pet being part of himself; a member of the family who goes through the same

67

experiences as he. Thus, the way a child handles his pet is more expressive and revealing of his

attitude toward the world than [play therapy]… The pet provides him with an outlet for his

emotions and a way of abreacting to them” (p. 696). Again, Levinson promoted the role of the

companion animal in allowing the client to share and explore important aspects of his or her life,

and the role of the therapist in understanding these dynamics as crucial to creating corrective

emotional experiences (See Figure 8).

Parish-Plass (2008) cited ways in which a pet could serve as an attachment figure in

children with insecure attachment, using classic concepts within attachment theory, such being a

target for proximity maintenance, a safe haven, ameliorating distress, and as a secure base to

develop autonomy. She concluded that a relationship with a pet can allow the owner to

acknowledge and work through attachment issues. She also noted that, in observing others with

pets, a child with an insecure attachment could begin to learn and internalize social rules, natural

reactions to various behaviors, acceptance of interpersonal relationship responsibilities, and

deciphering nonverbal language.

Margolies (1999) also addressed the developmental role that pet ownership could have,

particularly for women who have histories of insecure, anxious, or disorganized attachments. “A

primary attachment relationship with a pet offers the provision of security with less fear of

abandonment than might be experienced with an adult peer” (p. 297). Additionally, pet

ownership can be dialectically healing, as “pets can serve as both a source and an object of

nurturance” (p. 297). In situations in which an individual has experienced an inconsistent or

severed attachment, a pet could serve as a bridge to a stronger therapist-client relationship by allowing for tangible discussions of trust, positive regard, and such discourse could serve as scaffolding for other interpersonal relationships in the client’s life.

68

Jessee (1982) spoke to the nature of pets, and their role in enriching the daily lives of the humans around them, noting "Small animals, with their spontaneity and eagerness for affection, encourage many people to open up for the first time in their hospitalization. Animals often serve as a nonthreatening bridge to the present world for the person who has lost touch with reality and is unresponsive to staff interventions...Small animals help patients who are preoccupied or self- conscious about physical limitations to come out of their shells and interact. Pets do not reject someone who is an amputee, nor do they look away because a person is old and in a wheelchair, and they do not avoid the person who cannot speak clearly. Animals provide an accepting relationship, since they seek attention from everyone" (p. 27).

Grief and loss. In her personal account of pet loss, Noonan (2009) sought to define the connection between human beings and their pets. After describing her bond with her cat, she traced the history of pet ownership in the human experience, then discussed theoretical conceptualizations of the HAB, citing research that pet owners have larger affiliative needs, are more dependable, and are more responsible than individuals who do not love or own pets. She did not provide empirical evidence for these claims, however. Noonan also compared the pet- human bond to infant-adult attachment, defining an attachment figure as “someone who reliably provides comfort and protection and who offers a secure base from which the infant can begin to explore the environment, physically and emotionally” (p. 399). This type of relationship is internalized into a in the future, and Noonan suggested that pet ownership can provide a clear example of secure attachment. She further conjectured that the pet may at times serve as the secure base or caregiver figure to the human owner.

Kwong and Bartholomew (2011) analyzed the grief experienced by individuals who had lost an assistance dog using the lenses of attachment and caregiving. They noted that the

69

connection between a person with a disability and his or her service dog is multi-faceted: more time is spent together as a part of training; the dog has both instrumental and pet-like functions; and the dog is likely to accompany the owner to a wider variety of environments than a pet dog would. Kwong and Bartholomew conducted interviews with 25 individuals who had experienced the loss of at least one assistance dog. They asked whether the attachment components of safe haven, secure base, proximity-seeking, and separation anxiety were present in the relationship, and worked to illuminate if and how such processes were experienced by the owner. They also queried for components of caregiving, and explored the grief experiences of each participant following the death of the assistance dog. Deductive and inductive thematic analyses revealed that 21 of the 25 participants characterized their dogs as providing a safe haven, and 13 described their dogs as secure bases for exploration. The attachment components of proximity seeking and separation anxiety were less clearly enunciated, with most of the participants stated that they

were rarely if ever separated from their dogs, and clarified that they were typically in very close

proximity to one another. In terms of caregiving, 23 of the 25 participants described their

relationships with their dogs as being reciprocally caregiving, meaning that both the dog and the

owner took care of another and seemed to find the arrangement mutually beneficial. In their

exploration of the grief experienced by the owners, participants compared the loss to the death of

a family member, romantic partner, close friend, or a body part. Kwong and Bartholomew

concluded that, in terms of providing a secure base for exploration and a safe haven in times of

distress, assistance dogs were indeed sources of attachment. They also noted that, in this sample,

the HAB seemed to be best described in terms of attachment theory and reciprocal caregiving

(2011).

70

Figure 8. Comparing Types of Bonds.

Affectional Bond •Persistent Over Time •Specific Between Two Living Things •Emotionally Significant Attachment Bond •Affectional Bond •Secure Base •Safe Haven •Proximity-Seeking •Separation Anxiety Human-Animal Bond •Affectional Bond •Secure Base •Safe Haven •Unconditional Love

Figure 8. This graphic depicts the forms of connections discussed throughout the section: affectional bonds, which have been researched between various living organisms, attachment bonds, which have primarily been researched between humans, and aspects of the human-animal bond.

Operationalizing Pet Attachment

Ainsworth’s Strange Situation Procedure. As previously noted, gender has been indicated as having a significant impact on the HAB physiologically in the form of oxytocin levels (Miller et al., 2009). Prato-Previde, Fallani, and Valsecchi (2006) sought to determine whether gender had an impact on behavioral interactions between pets and owners. Using the assumption that the human-dog relationship could be likened to attachment behavior between caregivers and infants, they recruited 25 men and women to participate in a modification of

Ainsworth’s Strange Situation Procedure (ASSP), observing vocal and non-vocal interactions during separation, introduction of a stranger, reunion, and play. Participants also completed the

Pet Attachment Survey Questionnaire (PAS) to determine the relationship between the owners’

71

observed behaviors and their self-reported attachment to their dogs. Consistent with their

hypotheses, gender had a significant impact on vocal behavior, with female participants talking

to their dogs more quickly, with more words, terms of endearment, baby-talk, and for longer

periods of time than men. Gender was not significant in nonverbal interactions, such as

comforting, petting, and playing, though women had “to trouble more in controlling their dogs”

(p. 68) when leaving the room. On the PAS, the subscales of relationship maintenance and

intimacy were moderately correlated (ρ= 0.42, p=0.038), but no significant differences were

found between men and women. One limitation of this well-controlled study was the decision to

use the standardized protocol for the ASSP. While this is a sensible option, given the numerous replications of results in attachment studies, one cannot control for order effects of introducing strangers or owners, which could confound the observations. In particular, some of the behavioral changes could have been the result of fatigue, as opposed to the absence of the dog owner/attachment figure (Prato-Previde, Fallani, & Valsecchi, 2006).

Parthasarathy and Crowell-Davis (2006) built on previous research about the canine

perspective of attachment, which had found that dogs prioritized human relationships studies

(Scott & Fuller, 1965; Tuber et al., 1996) and that specific attachment patterns exist between

dogs and owners (Topal et al., 1998). They investigated dogs with and without separation

anxiety in an effort to determine whether the disorder was associated with problematic

attachment to the human owner, and found that dogs with and without separation anxiety spent a

similar amount of time in close proximity to their owners during the assessment. Both groups of

dogs also spent about the same amount of time near the exit doors, engaging in anxiety-related behaviors, and engaging in proximity-seeking behaviors when videotaped at home. Dogs with separation anxiety did, however, spend significantly more time jumping on the door after the

72

stranger left the room, and they spent significantly more time stationary when their owners were

in the room in the assessment. The authors concluded that separation anxiety does not stem from

hyperattachment to a human caregiver. Unfortunately, the human participants were not assessed

in any way, making it difficult to determine the presence or degree of the HAB as perceived by

the pet owner. Additionally, the ASSP was applied in the same order to all participants, so the authors were unable to control for potential order effects in the observed results (Parthasarathy &

Crowell-Davis, 2006). Despite this, the study offered evidence that attachment bonds and behaviors exist between dogs and their owners.

Palmer and Custance (2008) sought to address the potential impact of order effects by counterbalancing the procedure in the ASSP. They recruited 38 human-pet pairs (33 women, 5 men, 19 female dogs, 19 male dogs) for the procedure, and randomly assigned them to one of two conditions: the classic ASSP, where the owner and dog were placed in the novel environment together first, and a modified, counterbalanced ASSP, in which the dog was placed in a novel environment with a stranger first. Prato-Previde et al.’s (2003) results were replicated, in that the dogs exhibited search, proximity-seeking, and comfort-seeking behavior throughout the ASSP. Extending those results, however, Palmer and Custance found that dogs also exhibited exploration of their environments when in the presence of their owners, indicating that owners served as a secure-base. When dogs were placed in the unfamiliar room with their owners first, they displayed significantly more exploration behaviors than the dogs that were placed in the room with a stranger. Over time, exploration levels decreased in both conditions, as has been observed with human infants. In contrast to human iterations of the ASSP and regardless of the condition, dogs did not exhibit fear or discomfort about the stranger and instead tended to approach or ignore the person. Dogs in the first condition did, however, approached the stranger

73

significantly more that dogs in the second condition, leading the authors to conclude that this was

a manifestation of the secure-base effect. Consistent with infant-caregiver results, dogs explored less, were more passive, and engaged in less independent play when their owners were not present. Palmer and Custance (2008) did not consider the effect of owner gender in the study, and provided no explanation for this oversight, which could have provided additional information, given that other studies indicated that men and women behave differently with their animals (Prato-Previde, Fallani, & Valsecchi, 2006). Again this study offered evidence that the

HAB can be measured using an attachment behavior procedure.

Topal and colleagues (2005) applied the Ainsworth Strange Situation Procedure (ASSP) to three groups of animals: 11 hand-reared dog puppies, 13 hand-reared wolf puppies, and 11 pet dog puppies. Their goal was to determine whether attachment behaviors could be observed, differential socialization would have an effect, and whether attachment differences could be measured across species. The wolf puppies spent the majority of the assessment exploring and playing, and seemed to prefer physical contact with strangers, as opposed to the human who hand-raised them. Hand-reared dog puppies and the pet dog puppies were nearly

indistinguishable from one another, and spent much time playing and exploring. In contrast to the wolf puppies, however, both groups spent more time in proximity to their owner than with a stranger. This study was significant in showing that the relationship between pets and humans is reflective of a unique type of relationship, and this may be due to attachment bonds between pet and owner (Topal et al, 2005).

Constructing self-report measures. Albert and Bulcroft (1988) sought to understand and operationally define pet attachment as a construct, and once this was completed, determine trends between pet attachment, family characteristics, and family life cycle stage. They

74

conducted telephone interviews with 612 pet owners and 251 non pet owners, representative of

both urban and rural households in the sample. Information was gathered about marital status,

number of children present in the home, stage in the family cycle, age, and income are related to

pet ownership. Factor analysis and inter-item correlations revealed that nine items which could be defined as pet attachment. Once Albert and Bancroft (1988) crafted the nine-item scale, telephone interviews were coded, revealing several trends. Companion animals were viewed as family members by people who live in urban areas and served as emotional and psychological helpers with that population. Additionally, owning a pet and the perception of companion animal roles varied depending on the respondents’ developmental stage and family type. Fewer families with young children owned a pet, and the authors suggested that the time-intensive nature of raising infants and toddlers could result in pets being an additional stressor, rather than a helpful or soothing life component. In families with children who were in elementary, middle and high school, however, pet ownership was high, while attachment (as defined by the scale) was low.

Albert and Bancroft stated this could be a result of using pets as teaching tools for children of this age range. A low percentage of individuals classified as both widowed and low-income suggested that these individuals may believe the financial burden of having a companion animal is greater than the experience of owning a pet. Pets were considered to be important figures for affection and attachment among divorced, never-married, and widowed people, as well as childless couples, newlyweds, and “empty-nesters”. The authors discussed that this finding could mean that pets can serve as emotional surrogates for family members throughout the life cycle.

In a study similar to Allen and Bulcroft’s (1988), Cox (1993) assessed the family adaptability, cohesion, and strength of the HAB in 116 families using the Family Adaptability and Cohesion Evaluation Scale (FACES-III) and the Pet Attachment Scale (PAS). She found that

75

family pet attachment accounted for 6% of the total variance in family adaptability (r = 0.25, p <

0.01) and 2.5% of the total variance in family cohesion (r = 0.16, p < 0.05).

Krause-Parello (2012) published a cross-sectional study of older women using a number

of variables: loneliness as quantified using the Revised UCLA Loneliness Scale; pet attachment

as defined by the Pet Attachment Scale (PAS); human social support as measured by the Seeking

Support subscale of the Coping Strategy Indicator (CSI); and depressed mood using the

Depressed Mood subscale of the Psychological General Well-Being (PGWB) Schedule. The convenience sample was comprised of 159 women between the ages of 55 and 84 years, was

97% Caucasian, and 45% widowed. Participants completed the aforementioned assessments, and subsequent analyses indicated that pet attachment support acted as a coping resource, mediating the relationship between loneliness and depressed mood. An inverse relationship was found between depressed mood and loneliness, as well as depressed mood and pet attachment. No significant relationship emerged between human social support and depressed mood was found, suggesting that pet attachment served as a greater source of help in the observed sample.

Unfortunately, as with other published studies, the sample was largely homogeneous in nature, with regard to participants’ racial, socioeconomic, educational, and lifestyle backgrounds.

Additionally, Krause-Parello did not specify whether the scales selected were reliable or valid for the sample. Some of the constructs were measured by as few as three items, which could have a deleterious effect on the results in the form of either a type I or type II statistical error in terms of significance (2012).

Beck and Madresh (2008) attempted to clarify the concept of attachment as it applies to companion animals, noting first that the term is often applied without theoretical connections. In critiquing published assessments, they noted that the Pet Attachment Scale (PAS; Albert &

76

Bulcroft, 1988) and the Lexington Attachment to Pets Scale (LAPS; Johnson, Garrity, &

Stallones, 1992) both appeared to capture attachment issues but did not explicitly link their items

to attachment theory. Research on the human attachment system originated in studies of

children's relations with caregivers but has since been generalized to investigate adult

relationships, particularly romantic partnerships, peer interactions, and connections. Beck

and Madresh (2008) argued, “A relationship need not be a true attachment in order to serve

‘attachment-related’functions” (p. 45). In their study, they compared relationships with humans

to relationships with pets in an endeavor to understand the differences between the two and to

connect research on the HAB with attachment theory in a systematic manner. They used

published scales to assess romantic attachment, and adapted those measures by substituting the

word “pet.” The scales used were the Relationship Questionnaire (RQ, Bartholomew & Horowitz

1991) and a shortened version of the Experiences in Close Relationships-Revised Scale (ECR-R;

Fraley,Waller & Brennan, 2000). The RQ consists of four prototypical attachment styles: Secure,

Preoccupied, Dismissing-Avoidant, and Fearful- Avoidant. The ECR-R contains two scales of

relationship insecurity: Avoidance and Anxiety. The study sample was comprised of 192 web

survey respondents (168 females and 24 males) ranging in age from 18 to 56 years. In the

sample, 29% owned cats, 36% owned dogs, and 35% owned both. Results were consistent with

theoretical predictions of attachment styles and relationships. Respondent age and gender were

not significantly correlated with any relationship measures. Pet and partners ratings were significantly correlated for the Preoccupied prototype (r = 0.36) and for the Dismissing-

Avoidant prototype (r = 0.26). Anxiety scales for pets and partners were also correlated (r = -

0.30). Married individuals reported lower insecurity levels than unmarried individuals on Partner

Avoidance and Anxiety, but the two groups did not differ on RQ prototype items, Pet Avoidance,

77

or Pet Anxiety. The number of dogs owned was inversely related to ratings for the Pet

Preoccupied prototype (r = -0.19) and Pet Anxiety (r = -0.28). Respondents who owned at least

one cat but no dogs had high scores on the partner Anxiety scale(r= 0.28). Respondents who

owned more dogs reported less romantic relationship anxiety and less avoidance with their pets.

Cat owners endorsed more partner Anxiety. Overall, pet relationships were more secure than

relationships with romantic partners, and on every scale, respondents endorsed more positive

relationship characteristics for their relationships with their pets. After finding that the structure

of pet and partner relationships is similar in their sample, Beck and Madresh correlated

corresponding pet and human scales to team whether these attachment styles or patterns were

consistent across relationship domains. Pet and partner versions of the Anxiety scale,

Preoccupied prototype, and Dismissing-Avoidant prototype were mildly correlated (r = 0.26 to

0.36), whereas the Avoidance Scale, Secure prototype, and Fearful-Avoidant prototype were uncorrelated. In general, however, their results were consistent with the hypothesis that romantic partnerships are impacted by prior romantic relationships, while pet relationships seemed to be defined differently (Beck & Madresh, 2008).

Beck and Madresh (2008) reported several conclusions and contributions to the literature with regards to applying attachment theory to pet ownership. Based on their findings, attachment scales created for adult relationships could be reliably and consistently adapted to reflect attachment relationships with pets. Another significant finding was a weak association of relationship styles when comparing romantic partners to pets, suggesting that these relationships could be based on different working models. They also noted that pet owners may experience more security in relationships with their pets than with their romantic partners. The authors did not, however, acknowledge the lack of diversity in their sample, as respondents were primarily

78

female and well-educated, and all had access to the internet in order to complete the web survey.

They acknowledged that additional research could focus more specifically on the way in which

secure attachment to pets could affect overall well-being, by comparing individuals who report secure relationships with pets to non-pet owners on measures of security in human relationships.

Woodward and Bauer (2007) focused on the HAB, the concept of attachment, and the application a circumplex model of interpersonal theory, which “describes state-like personality characteristics and provides a set of predictions about the impact different types of interpersonal behavior will have on the members of a dyadic interaction” (p. 171). Woodward and Bauer asserted that various behaviors have the ability to elicit predictable reactions from others in a way which maintain the dyadic system. Their study had two goals: to investigate whether the circumplex model of behavior could be generalized to the owner-pet dyad, and to use this model to predict owners’ impressions of the HAB. The sample was comprised of 266 undergraduates

(213 women, 53 men; 96% Caucasian) and 407 web-based respondents who reported that they currently owned at least one pet. Participants’ Interpersonal styles were measured using the

Impact Message Inventory-Generalized Others (IMI-GO). Pets’ interpersonal styles were assessed using an adapted version of the Impact Message Inventory- (IMI-SO).

These assessments yield eight scales and two general factors reflecting overall agency and affiliation. The HAB was measured utilizing the Pet Attachment Survey (PAS). Woodward and

Bauer created the Pet Behavior Scale (PBS) to evaluate the frequency and worth of positive pet behaviors. Results suggested that cats were viewed as significantly more hostile/distant than dogs, while dogs were rated as significantly more friendly-submissive than cats. Woodward and

Bauer (2007) also hypothesized that self-described dog and cat people would have different

79 interpersonal styles. The results indicated that those who preferred dogs were significantly less hostile than individuals who cited cats were their ideal pets (Woodward & Bauer, 2007).

Kurdek (2008) also created his own measure of pet attachment, consisting of four statements which were meant to capture the four components of secure attachment: safe haven

(“When I am feeling bad and need a boost, I turn to my dog to help me feel better”), secure base

(e.g., “I can count on my dog to be there for me”), proximity maintenance (e.g., “I like having my dog near me”), and separation distress (e.g., “I miss my dog when I am away from him or her”) (Kurdek, 2009, p. 441). He found that proximity maintenance was the most salient attachment concept for college students who lived with pet dogs, while safe haven was considered the least prominent. Scores of pet attachment moderated their perception of proximity to both human attachment figures and their dogs, such that participants with high levels of pet attachment reported being as close to their pets as they were to their mothers, , best friends, and significant others. These students indicated they were more connected to their dogs than their . The survey was not a validated or vetted assessment, though it had face validity as it specifically defined interaction with one’s dog in connection to formal attachment theory. Kurdek (2009) extended this work to adults in general, and focused on the construct of safe haven, a key component of classic attachment theory, by examining whether and how humans interact with their pet dogs in stressful situations. He measured the importance of safe haven in comparison to secure base, proximity maintenance, and separation distress. Although he hypothesized that safe haven would be the least impactful, he hoped to examine the relative importance of this in the lives of dog owners by determining the likelihood that respondents would seek their pets in stressful times over other attachment figures (mothers, fathers, brothers, sisters, friends, romantic partners, or children). Kurdek hypothesized that participants with

80

specific traits would be more likely to select the dog as safe haven over other attachment figures:

males, widowed individuals, persons who are highly involved in the caretaking of the dog, and

individuals who have low rates of self-disclosure.

Kurdek’s 2009 sample was comprised of 975 individuals (789 females and 186 males)

between the ages of 19 to 82 years. Most of the participants were Caucasian (96%), employed

full-time (61%), married, (64%), and had some college education at the minimum (92%). No

validated measure of safe haven was available for use with pets and humans. Kurdek adapted the

10-item Emotional Reliance scale for use in this study using a pilot sample of 121 college

students rated dogs, mothers, fathers, brothers, sisters, best friends, and romantic partners on the

original scale to identify a smaller, meaningful set of items. This scale involved rating how true it

was that the respondent would go to each figure when experiencing positive or negative feelings.

The resulting measure had four items: “I turn to my dog when I am alone or depressed,” “I turn

to my dog when I am feeling very bad about myself and need a boost,” “I turn to my dog when I

am feeling overwhelmed by responsibilities and commitments,” and “I turn to my dog when I am

disappointed” (p. 441). To assess pet attachment level, he used the same four-item questionnaire

that had been constructed for the 2008 college student study, and conducted a confirmatory

factor analysis in which each item loaded on its own factor but was allowed to correlate to the

other factors. This model was a good fit, χ2 (89, N = 975) = 527.76, p < .01. He also noted the four-factor model was a better fit than a one-factor model, χ2 (6, N=974) =858.94, p< .01.

Results of Kurdek’s 2009 study indicated that proximity maintenance was the most

salient attachment concept for adults with dogs, while safe haven was considered the least

prominent. A two-level hierarchical linear regression in which attachment figures were in level

one and participants were in levels two indicated that dogs were considered a safe haven more

81

often than mothers, fathers, brothers, sisters, best friends, and children. Partners were the only

attachment figures considered more of a safe haven than dogs. Kurdek’s final hypothesis, that

participants possessing specific characteristics would be more likely to select the dog as safe

haven over other attachment figures, was supported. In the area of gender, he found that men

perceived dogs as safe havens over mothers, sisters, best friends, and children more strongly than

did women, but preferred partners over dogs more strongly than did women. In the area of

marital status, he found that individuals who were single or cohabitating rated dogs as equivalent

to partners, married participants rated partners higher than dogs, and respondents who were

divorced or widowed rated dogs more highly than partners. Individuals who were primary

caretakers for dogs were significantly more likely to rate dogs as safe havens more than human

attachment figures. Individuals with low levels of self-disclosure in general were particularly

likely to rate dogs more favorably as safe havens than mothers, fathers, brothers, sisters, and

children, but rated dogs and partners equivalently. In listing limitations, Kurdek acknowledged

that his sample may have been skewed, as the participants were highly involved with their pets

and evaluated them very highly. He also noted that participants rated multiple attachment figures

in a standard rather than a randomized manner, sharing that order effects could have affected

these ratings. Beyond this, the study used self-report measures, in contrast to observational or interview formats more typical of attachment theory application.

Winefield and colleagues (2008) also created their own pet attachment measure, and validated it on an older sample of Australian adults over the age of 60 years. A total of 312 (130 males and 182 females) individuals agreed to participate provide information, with 179 being pet owners. The authors used an adapted form of the Multi-Dimensional Support Scale (MDSS;

Winefield, Winefield, & Tiggemann, 1992) to measure respondents’ satisfaction with emotional,

82 practical, and informational support from family and close friends and from health professionals.

The second measure, The Owner-Pet Relationship (OPR) questionnaire, consisted of items that focused on the owner’s wish to be in close proximity with his or her pet and perception of the bond as emotionally supportive and reciprocal. While the authors stated their questions were grounded in traditional attachment theory, the four constructs of safe haven, proximity-seeking, secure base, and separation distress were not explicitly included in the measure. The researchers measured health outcome with the SF-36 Health Survey, which features the following subscales: physical functioning, disability due to physical problems, physiological pain, general health perceptions, vitality, social functioning, disability due to emotional problems, and overall mental health. After controlling for age, sex, smoking, exercise, and perceived support from confidants and health professionals, neither pet ownership nor pet attachment added anything significant to the explained variance in health scale scores. Unfortunately, the Owner-Pet Relationship questionnaire was not validated prior to administration, and did not encompass the theoretical construct of attachment, which limits the practical meaning of the results.

Wells (2009) contributed to the discussion about the definitions and spirit behind the concepts of the human-animal bond (HAB) and attachment, observing that it may be more practical to investigate the types of support facilitated by relationships between humans and their pets, as opposed to trying to fit the relationship into the traditional framework of attachment. She noted that social support could be viewed as the sense that one is loved, valued, and has a place within his or her social network. Through this lens, pets provide social support by their outward behaviors as well as the owner’s perception that the companion animal is nonjudgmental and fully present. Wells noted that such support could have a health benefit regardless of whether it was merely perceived or both perceived and provided, citing human relationship evidence

83

(Pierce, Baldwin, & Lydon, 1997; Ratnasingam & Bishop, 2007; Smith, Ruiz, & Uchino, 2004);

that “perceptions of social support may be a more important correlate of health outcomes than

actual support… thus, the mental representations of support that owners have of their pets may

explain some of the health benefits witnessed in this population” (p. 531).

Current trends in research. A more recent study (Beetz et al., 2011) focused on

attachment theory and stress regulation, and the possible connection to pet ownership. They

noted that dogs could serve as secure attachment figures because they provide a secure base for

exploration, a place of safety in times of distress, their physical presence could elicit positive

emotions, and separation from them can be associated with pain or anxiety. The authors

hypothesized that dogs would have a more significant stress-reducing effect than humans in children with insecure/disorganized attachment when they were exposed to a social stressor.

They noted that focusing on individuals with insecure and disorganized attachment was clinically relevant, because such individuals often experience intense difficulty establishing alliances with counselors or teachers. The authors further speculated that if dogs could prove effective in novel or stressful situations, they could serve as a bridge for future secure or healthy connections with others. A second goal of the research project was to measure stress modulation (using salivary cortisol levels) at five times during a standardized social stress test in children with insecure/disorganized attachment in the presence of a real dog, a friendly human, and a toy dog

(control). Participants were 31 boys in second, third, and fourth grades from schools in Germany and Austria. They found that interaction with a real dog lowered salivary cortisol levels significantly more than interaction with a toy dog or a friendly female student during a challenging social situation. The more children stroked, and had body contact with the dog the lower were their cortisol levels after the experiment, suggesting that physical contact, in addition

84

to mere presence, may have an additive effect on reducing stress levels. Self-report measures of distress level indicated no significant differences between the three conditions. The participants stated they experienced a minor level of arousal overall, and shared that they were experiencing a positive mood both before and after the stressful social situation. The authors cited previous studies which indicate that persons who exhibit an insecure-avoidant attachment style tend to disregard or underreport unpleasant feelings when they encounter laboratory-induced stressors.

Results indicated that the children had trusting attachment relationships with their pets, irrespective of their human attachment, sharing male children with highly insecure/disorganized attachment appear open to relate to a dog, and therefore are able to utilize its presence for social support and stress reduction to a greater degree than a friendly adult (Beetz et al., 2011).

Zilcha-Mano, Mikulincer, and Shaver (2011b) cited other research which has successfully applied attachment theory to relationships with God, specific locations (such as a familiar home or community environment), and inanimate objects, then provided a framework for extending the HAB by formal application of attachment theory to animal-assisted therapy.

They also acknowledged that attachment style can change, subtly or dramatically, depending on one’s environment and life experiences. Based on this, they decided to compare attachment within human–pet relationships to those of couple relationships and child–parent relationships by organizing the relationship in terms of attachment anxiety and avoidance. Zilcha-Mano,

Mikulincer, and Shaver (2011a) constructed and validated the Pet Attachment Questionnaire

(PAQ), a self-report measure which assesses both major dimensions of attachment insecurity in human–pet relationships. The avoidance subscale assesses the level of discomfort owners experience when they are physically or emotionally close to their pets and the degree to which they work to maintain emotional or physical distance from their pet. The anxiety subscale

85 measures the intensity and intrusiveness of worry that something bad could befall the pet and self- about how much the pet values the owner. A low score on both subscales indicates a secure attachment. In validating the instrument, the researchers found that high levels of anxious pet attachment were correlated to high levels of psychological distress and low levels of psychological well-being, as measured by the Mental Health Inventory (MHI, Hebrew version).

They also found a significant correlation between people’s attachment orientations in human–pet relationships (as measured by the PAQ) and their attachment orientations in human– human relationships (as measured by the Experiences in Close Relationships Scale, ECR,

Hebrew version). Specifically, they found a direct relationship between anxiety and/or avoidance scores in human–human relationships and the same scale scores in human-pet relationships.

Human-human attachment anxiety was associated with both pet anxiety and avoidance, while human-human avoidance was significantly associated only with pet attachment anxiety, suggesting that, while some comparisons between interpersonal attachment and human-pet attachment may be made, the overlap between the two constructs is incomplete (Zilcha-Mano,

Mikulincer, & Shaver, 2011a). Again, this study relied on self-report information, as opposed to including the AAI or ASSP for additional verification and authentication of the findings.

Zilcha-Mano, Mikulincer, and Shaver (2012) built on their 2011 research, which had suggested that pets could serve as attachment figures, with two additional studies. They hypothesized that proximity to a pet would have beneficial effects on a bonded owner during a stress-inducing task (indicating that the animal could serve as a safe haven) or in an exploration task (suggesting that the pet was a secure base). The studies were similar in that both involved administration of the Pet Attachment Questionnaire (PAQ, to quantify pet attachment orientation) and the Experiences in Close Relationships Scale (ECRS, to quantify human

86

attachment orientation); took place in the participants' homes; involved random assignment to

one of three conditions (the pet was present in the room and the participant was asked to think

about it, the pet was not present in the room and the participant was asked to think about it, or the

pet was not present in the room and the participant was not asked to think about it). In the first

study,165 participants were asked to generate personal goals and rate their confidence in being

able to accomplish the goals. This was an exploration task, and the researchers postulated that

this would be a way for participants to view their pets as a secure base. Results indicated that

participants who thought about their pet (regardless of whether the pet was in the room) listed

significantly more goals, and expressed more confidence in their ability to attain these goals,

than participants who were in the control condition [F(2, 162)= 9.05, p<.01 eta2=.10 for number

of goals and (2, 162)= 5.02, p<.01 eta2=.06 for confidence]. Upon further analysis using human

attachment orientation and pet attachment orientation as covariates, they found that pet

attachment orientation moderated these effects, meaning that the increased number of goals and

enhanced confidence was present when participants were securely attached to their pets

(meaning their PAQ scores were low in attachment avoidance and attachment anxiety). In the

second study, 120 participants were asked to complete a challenging cognitive task, then rate the

threat level or difficulty of the task. Throughout the experiment systolic and diastolic blood

pressure was measured. This was a distress-eliciting task, and the researchers postulated that this would be a way for participants to view their pets as a safe haven. Results indicated that

participants who thought about their pet (regardless of whether the pet was in the room)

experienced a lower increase in blood pressure and were less likely to appraise the task as

threatening than participants who were in the control condition. Upon further analysis using

human attachment orientation and pet attachment orientation as covariates, they found that pet

87 attachment orientation moderated these effects, meaning that the buffered blood pressure increase and task appraisal dampening was present when participants were securely attached to their pets (meaning their PAQ scores were low in attachment avoidance and attachment anxiety).

Zilcha-Mano, Mikulincer, and Shaver (2012) concluded that in their sample, pets could fulfill the role of being a safe haven and a secure base when the pet owner identified the relationship as being securely attached. The study design was significant in that participants' attachment orientations were measured and classified, then included as a level of analysis, in comparison to the presence or absence of the pet in the room. By measuring the quality and type of attachment between the pet and the person, the results strengthen the argument that pet attachment is a valid construct, and lend credence to applying attachment theory to the HAB.

Summary

Humans and animals have coexisted for thousands of years, and archaeological findings suggest that animals have cohabitated and served as companions for 12,000 years (Morrison,

2007). The HAB has been theorized and explored, and numerous researchers have found evidence that companion animals have beneficial physiological effects on their human caregivers, in terms of lowering blood pressure (Friedmann et al., 1980; Friednmann, 1990), reducing one's reliance on pain medication (Lust et al., 1990), and anxiety reduction (Friedmann et al., 1983). Researchers have also found support for psychosocial benefits from the HAB, using case studies (Davis, 1988; Kovacs et al., 2004; Sellers, 2006; Sockalingham et al., 2008; Zimolag

& Krupa, 2010), qualitative interviews (Cline, 2010; Cohen, 2002; Knight & Edwards, 2008;

Yorke et al., 2008), personality assessments (Bagley & Gonsman, 2005; Johnson & Rule, 1991), and surveys (Antonacopoulos & Pychyl, 2010; Taylor & Sginal, 2005). Counselors and therapists have also reported using their own pets as therapeutic adjuncts, often with anecdotal

88 success (Draper et al., 1990; Levinson, 1965; Margolies, 1999; Mason & Hagan, 1999; Walsh,

2009).

Attachment theory is widely recognized and applied to the exploration of human interpersonal relationships. As with the HAB, researchers have traced neurobiological and physiological evidence of attachment in human development (Beatson & Taryan, 2003; Luyten

& Strathearn, 2011; Schore, 2001; Siegel, 2001), as well as documented health outcomes related to attachment styles and orientation (Sadava et al., 2009; Siegel, 2001; Strathearn, 2011). Cohen

& Wills (1985) conducted an extensive literature review and meta-analysis to determine whether the oft-cited connection between well-being and social support was the result of a direct-effect model or a buffering model. They found that the buffering model was correct when assessments of social support queried for the perceived availability of support in times of stress, suggesting that the presence and availability of a support system reduced the negative impact of distress.

They found evidence for a direct effect model when the social support measures assessed the degree to which a respondent perceived he or she was integrated within a social network, indicating that when a person believed he or she was a part of a social network, that person endorsed having a greater sense of well-being. Their work suggests that perceived social support can be a reliable indicator of subjective well-being for individuals both in times of distress and in general.

Multiple authors have used the term attachment to describe the connection between humans and their animal companions, assuming the constructs are interchangeable

(Antonacopoulos & Pychyl, 2010; Geist, 2011; Rynearson, 1978; Sacks, 2008). Such rationalizations are tenuous at best. Other researchers have used attachment theory tenets as a lens by which to understand the HAB, building a theoretical rationale prior to using terminology

89 such as secure attachment, secure base, safe haven, proximity-seeking, and separation anxiety

(Crawford et al., 2006; Kurdek, 2009; Kwong & Bartholomew, 2011; Zilcha-Mano et al., 2011,

2012). Many of these researchers have gone on to construct measures which accurately capture attachment concepts as they uniquely apply to pet ownership (Beck & Madresh, 2008; Kurdek,

2009; Zilcha-Mano, 2011). Others have applied traditional to the human- pet relationship (Beetz et al., 2011; Topal et al., 2005).

As Yorke (2010) noted, “Trust, safety, and structured interaction are important in animal companionship. Animal research confirms that all of these same ingredients are essential to normal and healthy development among all mammals” (p. 559). Given the ubiquitous nature of pet ownership in United States households, as well as throughout the world, it seems sensible to utilize an already-present form of attachment in a therapeutic environment. Indeed, the goals of psychotherapy include the establishment of trust and safety within the structured interactions between the therapist and the client. Introducing a companion animal to treatment could have a number of significant benefits.

Gaps in Research

Previous researchers have established that perceived social support and subjective well- being are related concepts, and suggested that a causal relationship may exist between perceived social support and well-being, such that enhanced social support may lead to a greater sense of subjective well-being in adults. As early as the 1970s, researchers began to focus on the role of social support, loosely defined as interpersonal transactions between at least two individuals, as a coping strategy for individuals. Conventional wisdom has held that social support has a positive effect on a person's well-being and may serve to protect him or her from psychological or physiological distress (Zimet et al., 1988). Although anecdotal evidence is helpful as a researcher

90

begins, theoretical frameworks and empirical evidence are far more effective in broadening one's

understanding. Some researchers have successfully applied attachment theory in their studies on

relationships, social support, and overall well-being (Sadava et al., 2009). Additionally, companion animals have come to occupy roles very similar to human beings over the past 75 years, and pets have become important sources of social support for many individuals

(Antonacopoulos & Pychyl, 2010). It is important to include pets in exploring social support as a psychological concept (Wells, 2009). This has not yet been done in a systematic, comprehensive manner, yet it could have a significant role in understanding why individuals own pets, what the human-animal bond means from an attachment perspective, and what benefits the pets provide to their owners in terms of overall health or as members of a support network. Reliable, valid instrumentation is crucial in establishing this connection.

Cohen and Wills (1985) conducted an extensive literature review to determine whether the oft-cited connection between well-being and social support was the result of a direct-effect model or a buffering model. They found that the buffering model was correct when assessments of social support queried for the perceived availability of support in times of stress, suggesting that the presence and availability of a support system reduced the negative impact of distress.

They found evidence for a direct effect model when the social support measures assessed the degree to which a respondent perceived he or she was integrated within a social network, indicating that when a person believed he or she was a part of a social network, that person endorsed having a greater sense of well-being. Their work suggests that perceived social support can be a reliable indicator of subjective well-being for individuals both in times of distress and in general. As Zimet and colleagues (1988) explain, "Although social support may be directly helpful in all circumstances, it may be particularly effective as a buffer during times of stress" (p.

91

31). This suggests that it is important to make sure that the measure being used to assess

perceived social support is theoretically sound, face-valid, and reliable. Zimet et al.'s 12-item

Multidimensional Scale of Perceived Social Support (MSPSS) fulfills these requirements and has been used and validated across multiple studies. It has not, however, been used as part of an assessment of attachment style, well-being, social support, and pet attachment (Zimet et al.,

1988). This study is one of the reasons that partial mediation is hypothesized.

Evidence also exists that attachment relationships can provide extensive social support

(Mikulincer & Shaver, 2007). Perceived social support from a romantic partner has been

consistently associated with an increase in one’s perceived ability to cope with stress, reducing

anxiety and rumination, and a significant reduction in one’s experience of pain (Cohen &

McKay, 1984; MacDonald & Leary, 2005; Thoits, 1986). Within the realm of secure attachment

orientations, researchers have found consistent links between relationship security, perceived

social support, and elevated mood (enhanced well-being). In adults, even the symbolic availability of attachment figures, through priming of attachment figure representations or asking a participant to visualize an attachment figure, has a soothing effect (Mikulincer & Shaver,

2007). Thoits (1986) suggested that the mechanism involved in social support is through assistance with coping. He clarified that a support system may help by ameliorating the stressful situation directly, softening the meaning or impact of the situation, or changing the person's emotional reaction to the situation. A companion animal may also be able to provide such assistance, but this area has not been empirically explored (Wells, 2009).

Ample experimental support exists for psychosocial and physiological benefits for the bond between humans and animals. Attachment theory has been successfully applied to relationships with God, specific locations (such as a familiar home or community environment), and inanimate

92

objects (Zilcha-Mano et al., 2011b). Numerous researchers, as noted above, agree that

attachment theory has multiple applications, including the relationship between pets and owners.

Zilcha-Mano and colleagues (2011) have constructed the Pet Attachment Questionnaire (PAQ), a self-report assessment which closely mirrors the Experiences in Close Relationships Scale (ECR) by Brennan and colleagues (1998), a benchmark assessment of adult attachment. Both scales gauge two factors, anxiety and avoidance. Pet attachment avoidance has been likened to discomfort owners experience when they are physically or emotionally close to their pets and the degree to which they work to maintain emotional or physical distance from their pet. Pet attachment anxiety refers to intensity and intrusiveness of worry that something bad could befall the pet and self-doubt about how much the pet values the owner. In an Israeli sample, Zilcha-

Mano and colleagues found a significant correlation between human–pet attachment orientation human–human attachment. The PAQ has recently been translated into English, and to date, no studies have been published regarding its validity, reliability, or practical utility in an English- speaking sample. In order to extend the research in anthrozoology, it will be important to use reliable, validated instruments to measure pet attachment as a covariate in empirical research.

Additionally, it could be beneficial to replicate and extend previous studies’ methods with validated, meaningful tools, and increase the generalizability of results by using diverse sampling, while acknowledging individual differences in relationships between humans and their pets.

Overview of the Present Study

The current study investigated the associations between pet attachment (PAtt), perceived social support (PSS), adult attachment (AAtt), and subjective well-being (SWB) among adult pet-owning individuals. This study addressed several research questions. In what ways are adult

93 attachment style, subjective well-being, and perceived social support related to one another? In what ways are pet attachment and adult attachment related to one another in individuals? Does pet attachment have an association with ratings of subjective well-being? Does perceived social support have an association with pet attachment style? Does pet attachment mediate or moderate the relationship between subjective well-being and perceived social support? Because little is known about the connections between pet attachment, perceived social support, and subjective well-being, I chose to test both mediation and moderation models (hypotheses 12 and 13). Based on these research questions, the following hypotheses were tested:

Hypotheses

Based on the aforementioned research questions, the following hypotheses were tested:

1. Consistent with previous literature, higher levels of secure adult attachment (measured as

low anxiety and low avoidance) will be associated with greater subjective well-being and

perceived social support. More specifically: lower levels of adult attachment anxiety will

be associated with greater subjective well-being; lower levels of adult attachment

avoidance will be associated with greater subjective well-being; lower levels of adult

attachment anxiety will be associated with greater perceived social support; and lower

levels of adult attachment avoidance will be associated with greater perceived social

support.

2. Consistent with previous literature, perceived social support will be positively correlated

with subjective well-being.

3. Consistent with previous literature, perceived social support will positively predict

subjective well-being. This was a confirmatory analysis to determine whether the current

sample was similar to previous studies in terms of perceived social support and subjective

94

well-being (Cohen & McKay, 1984; Deci et al., 2006; MacDonald & Leary, 2005;

Sadava et al., 2009; Thoits, 1986).

4. Adult attachment security will predict subjective well-being.

5. Perceived social support will predict adult attachment.

6. The association between perceived social support and subjective well-being will be

partially mediated by adult attachment.

7. Lower levels of perceived social support will strengthen the association between adult

attachment and subjective well-being.

The following exploratory hypotheses will be tested:

8. Pet attachment security will be positively correlated with adult attachment, such that pet

attachment anxiety will be positively correlated with adult attachment anxiety; and pet

attachment avoidance will be positively correlated with adult attachment avoidance.

9. Similar to adult attachment findings, higher levels of secure pet attachment (defined as

low anxiety and low avoidance) will be correlated with greater subjective well-being and

perceived social support. More specifically: lower levels of pet attachment anxiety will

be associated with greater subjective well-being; lower levels of pet attachment

avoidance will be associated with greater subjective well-being; lower levels of pet

attachment anxiety will be associated with greater perceived social support; and lower

levels of pet attachment avoidance will be associated with greater perceived social

support.

10. Pet attachment security will predict subjective well-being after controlling for adult

attachment.

11. Perceived social support will predict pet attachment after controlling for adult attachment.

95

12. The association between perceived social support and subjective well-being will be

partially mediated by pet attachment.

13. Lower levels of perceived social support will moderate the association between pet

attachment and subjective well-being after controlling for adult attachment.

Figure 9

Hypothesized Mediation Model of Perceived Social Support, Adult Attachment, and Subjective Well-being- Hypothesis 6.

AAttAnx SWB AAttAvo PA PSSTotal NA

SWLS

Notes: AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; and NA = Negative affect.

Figure 9. Hypothesized mediating effects relating adult attachment (AAtt), total perceived social support (PSS), and subjective well-being (SWB). Full mediation would be indicated if the AAtt style was a significant predictor of SWB when PSS was not included in the model (see test of Hypothesis 4), but not when PSS was included in the model. Partial mediation would be indicated if the AAtt style was a significant predictor of SWB when PSS was excluded from the model and was still a significant predictor, but reduced in size, when PSS was included in the model.

96

Figure 10

Hypothesized Moderation Model of Perceived Social Support, Adult Attachment, and Subjective Well-being- Hypothesis 7.

PSSTotal

SWB

AAttAnx PA AAttAvo NA

SWLS PSSTotal x AAtt

Notes: AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; and NA = Negative affect.

Figure 10. Hypothesized moderating effects relating adult attachment (AAtt), total

perceived social support (PSS), the interaction between AAtt and PSS, and subjective well-being

(SWB). Hypothesis 7 predicted that lower levels of PSS would strengthen the association between adult attachment (AAtt) and SWB.

Figure 11

Hypothesized Mediation Model of Perceived Social Support, Pet Attachment, and Subjective Well-being- Hypothesis 12. PAQAnx SWB PAQAvo PA

NA PSSTotal SWLS

Notes: PAttAnx= Pet Attachment Anxiety Subscale; PAttAvo= Pet Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; and NA = Negative affect.

97

Figure 11. Hypothesized mediating effects relating pet attachment (PAtt), total perceived social support (PSS), and subjective well-being (SWB). Full mediation would be indicated if the

PAtt style was a significant predictor of SWB when PSS was not included in the model (see test of Hypothesis 10), but not when PSS was included in the model. Partial mediation would be indicated if the PAtt style was a significant predictor of SWB when PSS was excluded from the model and was still a significant predictor, but reduced in size, when PSS was included in the model.

Figure 12

Hypothesized Moderation Model of Perceived Social Support, Pet Attachment, and Subjective Well-being- Hypothesis 13.

PSSTotal

SWB

PAQAnx PA PAQAvo NA

SWLS PSSTotal x PAQ

Notes: PAttAnx= Pet Attachment Anxiety Subscale; PAttAvo= Pet Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; and NA = Negative affect.

Figure 12. Hypothesized moderating effects relating pet attachment (PAtt), total perceived social support (PSS), the interaction between PAtt and PSS, and subjective well-being

(SWB). Hypothesis 13 predicted that lower levels of PSS would strengthen the association between PAtt and SWB.

98

CHAPTER III

RESEARCH DESIGN AND METHODOLOGY

Introduction

The study assessed the attitudes and perceptions of pet-owning adults. The sample, which

initially included 711 individuals, was recruited from email lists and social media engines using

a snowball method of recruitment (see Appendix B for recruitment flyer). Participants were then

provided with digitized informed consent forms and directed to complete the survey on Survey

Monkey. The survey was comprised of a demographics questionnaire (Appendix C), the

Experiences in Close Relationships Scale (ECR), The Pet Attachment Questionnaire (PAQ), the

Multidimensional Scale of Perceived Social Support (MSPSS), the Satisfaction With Life Scale

(SWLS), and the Positive and Negative Affect Scale (PANAS). The questions and scales were

randomly ordered in an effort to counteract any participant fatigue, and the entire survey took

approximately ten minutes to complete.

Incomplete surveys and surveys with outliers were removed, yielding a final sample size

of 561 pet-owning adults. Descriptive statistics were calculated, and all data analyses were

completed using IBM SPSS Statistics 21.

Instruments

Demographic Information Questionnaire. The Demographic Information Questionnaire

included items about age, gender, racial/ethnic background, geographic location, language

proficiency, educational level, relationship/marital status, religious affiliation, and

socioeconomic status (Appendix B).

Adult Attachment. In this study, adult attachment was examined using Brennan et al.’s

(1998) Experiences in Close Relationships Scale (ECR). As previously noted, the ECR is

99

comprised of two 18-item scales, one to assess attachment anxiety and a second scale to gauge

avoidant attachment. The scale includes twenty items rated on a 5-point Likert-like scale ranging

from 1 (strongly disagree) to 5 (strongly agree). Sample items include the following: (a) "I

worry about being abandoned," (b) "My desire to be close sometimes scares people away," and

(c) "I tell my partner just about everything." Across its use, the ECR has demonstrated high

reliability (α > .90 with test-retest coefficients between .50 and .75) and discriminant validity

(correlation between anxiety and avoidance subscales is close to zero). The validity of the ECR

has been tested across experimental manipulations and behavioral observation. An additional

benefit of this tool is the fact that the wording of the ECR has been successfully altered to

capture participants’ global attachment styles, their attachment styles in a specific romantic

relationship, and their retrospective attachment styles in other relationships. Additionally, the

ECR was used as a model for the construction of the Pet Attachment Questionnaire.

Pet Attachment. Zilcha-Mano, Mikulincer, and Shaver (2011a) constructed and validated

the Pet Attachment Questionnaire (PAQ), a 26-item self-report measure which assesses the anxiety and avoidance dimensions of attachment insecurity in human–pet relationships. A low score on both subscales indicates a secure attachment. Sample items include the following: (a) "

Often my pet is a nuisance to me," (b)"I have no problem parting with my pet for a long duration," and (c) "Signs of affection from my pet bolster my self-worth." The scale consists of items rated on a 7-point Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree).

In validating the instrument, the researchers found that high levels of anxious pet attachment were correlated to high levels of psychological distress and low levels of psychological well-being, as measured by the Mental Health Inventory (MHI, Hebrew version).

100

They also found a significant correlation between people’s attachment orientations in human–pet

relationships, as measured by the ECR. Adult attachment anxiety was associated with both pet

anxiety and avoidance, while human-human avoidance was significantly associated only with pet

attachment anxiety, suggesting that, while some comparisons between interpersonal attachment

and human-pet attachment may be made, the overlap between the two constructs is incomplete.

In evaluating the temporal stability and reliability of PAQ scores over the course of six months,

the authors found that Cronbach alphas were between .86 to .89, and the two scales were not

significantly correlated with one another (rs<.06). Pet attachment anxiety had a test-retest

reliability coefficient of .75, the pet attachment avoidance had a test-retest reliability coefficient

of .80 (Zilcha-Mano, Mikulincer, & Shaver, 2011a). Although previous studies have demonstrated the evidence of the validity of this instrument, a significant drawback is the lack of validity evidence with populations using the English version of the scale. Therefore, in the

present study, I attempted to provide evidence of the validity of the English translation of the Pet

Attachment Questionnaire using English-speaking adult samples.

Perceived Social Support. In 1988, Zimet and colleagues developed the Multidimensional

Scale of Perceived Social Support (MSPSS). This brief, 12-item self-report measure provides a subjective assessment of the adequacy of the respondent's social support from family, friends, and significant others using a 1 (strongly disagree) to 5 (strongly agree) Likert-style scale.

Sample items include the following: “There is a special person with whom I can share my joys and sorrows,” “I can talk about my problems with my family,” and “My friends really try to help

me.” It was validated on 275 university undergraduate students. In their creation and validation

of the measure, the researchers completed a confirmatory factor analysis and found that the items

had high loadings on three factors: perceived support from family, from friends, and from a

101

significant other ("special person"). There were intercorrelations among the subscales (r= .63 for

significant other and friend; r= .24 for significant other and family; r= .34 for family and friend).

Internal reliability for all three subscales was high, with Cronbach's coefficient alphas of .91 for

the significant other subscale, .87 for the family subscale, and .85 for the friends subscale. The

test-retest reliability was also high; 69 of the 275 original participants took the MSPSS two to

three months after the initial administration, with the following reliabilities: .72 for the

significant other subscale, .85 for the family subscale, .75 for the friends subscale, and .85 for the

whole scale overall (Zimet et al, 1988).

Satisfaction With Life. The Satisfaction With Life Scale (SWLS; Diener, Emmons,

Larsen, & Griffin, 1985) was used to assess life satisfaction. The SWLS items measure an individual’s global, cognitive evaluation of his or her satisfaction with life as a whole based on their own standards and values. The scale consists of five items rated on a 7-point Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree). Scores can range from 5 (low

satisfaction) to 35 (high satisfaction). Sample items include the following: (a) “If I could live my

life over, I would change almost nothing,” and (b) “The conditions of my life are excellent.”

Diener et al. (1985) reported a coefficient alpha of .87 for the scale and a two-month test-retest reliability coefficient of .82. The factorial validity of the SWLS has been replicated in several studies (Diener et al., 1985; Pavot, Diener, Colvin, & Sandvik, 1991). For instance, Diener et al.

(1985) concluded that a single factor accounted for 66% of the variance in the scale.

Positive and Negative Affect Schedule. The affective component of subjective well-being

was measured with the Positive and Negative Affect Schedule (PANAS; Watson, Clark, &

Tellegen, 1988). The PANAS has been used extensively in the literature on self-reported mood.

Specifically, it measures how much participants generally experience a particular positive (e.g.,

102

proud, interested, enthusiastic) or negative (e.g., nervous, irritable, distressed) emotion. The

scale includes twenty items rated on a 5-point Likert-like scale ranging from 1 (not at all) to 5

(extremely). The scales were constructed using a large sample of adults and undergraduate students. The alpha reliabilities of the scales ranged from .86 to .90 for positive affect and from

.84 to .87 for negative affect. The two factors accounted for two thirds of the common variance in mood, providing support for the factorial validity of the PANAS. The convergent validity of the scales was supported by significant correlations between the PANAS scales and measures of related constructs including distress, depression, and other brief affect measures. This 20-item measure of the major dimensions of mood appears to compare favorably to lengthier measures of these factors. The PANAS also permits researchers to designate various time instructions for participants self-reporting their mood.

Participants

The final sample consisted of 561 pet-owning adults who were at least 18 years of age.

The sample was recruited from list-servs, websites, and social media engines, with advertisements soliciting participation. More specifically, participants were recruited through the

American Psychological Association's Division 17 (Society of Counseling Psychology), 20

(Adult Development and Aging), 35 (Society for the Psychology of Women), and Division 44

(Society for the Psychological Study of Lesbian, Gay, Bisexual, and Transgender Issues).

Participation was also solicited through multiple human-animal interaction list-servs and the

Washington Counseling Association. In addition, a snowball method of recruitment was utilized by encouraging participants to request their pet-owning friends to participate in the study.

Participants who did not consider themselves to be proficient in reading English, as well as those

103

who did not currently own a pet, were excluded from the sample because they did not meet the

inclusion criteria of the study.

A total of 711 surveys were submitted electronically via SurveyMonkey, but 117 surveys

were excluded because the participant omitted age information (20 total surveys; this made it

impossible to verify that the participant was an adult at least 18 years of age) or because the

participant omitted two or more items within an instrument. In an attempt to remove potential

outliers, instrument subscale scores were converted to z-scores and all cases above 3 or below -3,

representative of being three standard deviations above or below the mean, were removed as

outliers. This resulted in the removal of 33 additional participant surveys due to extreme scores:

11 on the Pet Attachment Questionnaire (PAQ), 11 on the Positive and Negative Affect Scale

(PANAS), eight on the Multidimensional Scale of Perceived Social Support (MSPSS), and three

on the Satisfaction with Life Scale (SWLS). I discerned whether the data was normally

distributed using the K-S test and Shapiro-Welk test for normality. I also reviewed the skewness

and kurtosis of the data. Upon removal of the outliers, the data was found to be normally

distributed because it was within normal limits for skewness, kurtosis, and normality, as defined

by Field (2009) After this process, 561 participant surveys were used for final data analysis and hypothesis testing.

The final sample was comprised of 561 pet-owning adults, including 73 men (13.0%),

481 women (85.7%), one transgendered individual (0.2%), and six individuals who preferred not to answer or who identified gender in a different way (1.0%). Respondents ranged in age from

19-75 years, with a mean age of 22.71 years. More specifically, 1.1% were under 20 years old,

28.9% were 20-29 years of age, 29.2% were 30-39 years of age, 17.3% were 40-49 years of age,

12.5% were 50-59 years of age, 5.8% were 60-69 years of age, and 1.1% were over the age of 70

104

years. Five hundred four participants (89.9%) reported that their current country of residence was

the United States, and 492 respondents (87.7%) indicated that they were born in the United

States.

In the area of sexual orientation, respondents self-identified in the following ways: 83.2% were straight/heterosexual, 2.5% were gay, 6.2% were lesbian, 5.7% were bisexual, 0.4% were asexual, 0.7% were pansexual, and 2.7% preferred not to answer or identified in a different way.

In the area of relationship status, 17.0% of respondents reported being single, 48.3%

reported being married, 19.3% were cohabitating or living with a partner, 8.9% were in a

committed relationship but not living together, 5.3% were separated or divorced, and 0.7% were

widowed. One hundred seventy-two (30.7%) of the survey respondents reported being parents.

In describing the racial and ethnic identity of respondents, participants were able to select

more than one classification in which to identify. Because of this, percentages did not add to

100% in order to allow for a more accurate self-identification of racial and ethnic background.

Participants endorsed racial background in the following ways: 92.9% of the sample identified as

Caucasian, 0.9% identified as African-American, 2.7% identified as Asian-American, 1.1% identified as Native American, American Indian or Alaskan Native, 3.4% identified as Latino,

Hispanic, or Spanish-American, 0.5% identified as Asian-Indian, and 1.0% preferred not to answer. A total of 4.3% of respondents (n=24) reported having some Latino heritage.

The education level of participants was also assessed: 5.2% had attained at least a high school education, 5.7% had completed an associate’s degree, 25.3% has completed a bachelor’s degree, and 60.6% had completed a graduate degree.

Religious affiliation was also gauged: 38.1% identified as Christian, 23.3% reported not being religious, 10.4% were Agnostic, and 9.6% were Atheist. Fewer than 5% of respondents

105

selected one of the following categories: Unitarian/Universalist (3.7%), Preferred not to answer

(3.7%) Spiritualist (3.5%), Buddhist (2.8%), Jewish (2.8%), Humanist (1.5%), Muslim (0.4%),

Wiccan (0.2%), or Pagan (0.2%).

Respondents also provided information about the type and number of companion animals

they owned. When asked about dog ownership, 26.2% reported not owning a dog, 42.2% had

one dog, 21.6% had two dogs, and 9.8% reported owning three or more dogs. When cat

ownership was measured, 48.5% reported having no cats, 23.4% reported having one cat, 16.4%

had two cats, and 11.8% reported having three or more cats. Fifty-two participants (9.3% of the

sample) indicated horse ownership, 15 respondents (2.7%) reported that they had at least one pet

reptile, 20 individuals (3.6%) owned at least one rodent, 36 (6.4%) owned at least one fish, and

26 participants (4.6%) indicated owning at least one bird.

Survey participants were asked to indicate which pet they used to complete the Pet

Attachment Questionnaire (PAQ), because this instrument assesses the attachment bond between an adult and a specific companion animal. Fifteen individuals (2.7%) did not indicate the type of pet used, but the remaining participants indicated the following: 65.6% used a dog, 26.7% used a cat, 3.6% used a horse, and 4.1% used a different type of companion animal.

In summary, the majority of participants were women (85.7%), most resided in the

United States (89.9%), most identified themselves as being at least partially Caucasian (92.9%),

and most identified themselves as being straight or heterosexual in their sexual orientation

(83.2%). The sample was highly educated, with more than half of the survey respondents

indicating that they had earned a graduate degree (60.6%). Participants ranged in age from 19-75

years, with a mean age of 22.71 years. Respondents indicated an array of relationship statuses, as

well as religious affiliations. The majority of respondents owned at least one dog (74.8%),

106

approximately half owned at least one cat (51.5%), and various other animal types were reported

to be owned.

Procedure

The study was reviewed and approved by the Washington State University Institutional

Review Board (IRB) before dissemination of the materials. The sample was recruited from list-

servs, websites, and social media engines, with advertisements soliciting participation using an

email recruitment flyer (see Appendix B). More specifically, participants were recruited through

the American Psychological Association's Division 17 (Society of Counseling Psychology), 20

(Adult Development and Aging), 35 (Society for the Psychology of Women), and Division 44

(Society for the Psychological Study of Lesbian, Gay, Bisexual, and Transgender Issues).

Participation was also solicited through multiple human-animal interaction list-servs and the

Washington Counseling Association. In addition, a snowball method of recruitment was utilized

by encouraging participants to request their pet-owning friends to participate in the study.

Participants were provided with digitized written informed consent forms (see Appendix A)

describing the present study and were asked if they would be willing to participate. The

questionnaires were completed online via Survey Monkey.

Each participant was initially asked to complete a Demographic Information

Questionnaire (Appendix C) in order to obtain background information. Subsequently,

participants were asked to complete five questionnaires in English including (a) Brennan et al.'s

(1988) Experiences in Close Relationships Scale (ECR); (b) Zilcha-Mano et al.'s (2011) Pet

Attachment Questionnaire (PAQ); (c) 1988, Zimet and colleagues' (1988) Multidimensional

Scale of Perceived Social Support (MSPSS); (d) Diener et al.'s (1985) Satisfaction With Life

Scale (SWLS); and (e) Watson et al.'s (1988) Positive and Negative Affect Scale (PANAS).

107

Upon completing the demographic questionnaire, only participants who self-identified as having at least a basic degree of English proficiency and who currently owned a pet were asked to proceed with the remaining questionnaires. The order of the scales, as well as the order of the scale questions, were counterbalanced to counteract any possible fatigue issues. The questionnaires took approximately ten minutes to complete. Participants were allowed to contact the experimenter with any questions regarding the nature of the present study. Anonymity and confidentiality were maintained for all responses.

Data Analyses

Preliminary analyses. Descriptive statistics were reported for each measure. Pearson correlation coefficients were computed between all variables to determine the relationships between the constructs. Mediation and moderation hypotheses were tested using multiple regression analyses. SPSS-21 was used to conduct correlation and regression analyses for all hypotheses. All regression, mediation, and moderation analyses also included hierarchical regression, in which age, gender, relationship status, and educational attainment were entered into the first regression block in an attempt to control for these factors.

Hypotheses 1, 2, 8, and 9. To test hypotheses 1, 2, 8, and 9, correlational analyses

(Pearson’s r) were computed. I examined the Pearson correlations to determine whether the relationships between the variables were present, in the hypothesized directions, and significant at the p<.05 significance level using a one-tailed test of significance for each hypothesis.

Hypothesis 3. To test hypothesis 3, that perceived social support (PSS) would be a positive predictor of subjective well-being (SWB, as measured by the Multidimensional Scale of

Perceived Social Support and the Positive and Negative Affect Scale), linear regression analysis was used to determine whether perceived social support significantly predicted participants’

108 subjective well-being. This was a confirmatory analysis to determine whether the current sample was similar to previous studies in terms of perceived social support and subjective well-being

(Cohen & McKay, 1984; Deci et al., 2006; MacDonald & Leary, 2005; Sadava et al., 2009;

Thoits, 1986). The results of the regression analysis indicated the percentage of variance in subjective well-being scores explained by perceived social support, using the following equation: y = Xβ + ε and the following model: High PSS High SWB (high SWLS, high PA, low NA).

Hypothesis 4. To test hypothesis 4, that adult attachment style (AAtt) would predict subjective well-being (SWB), linear regression analysis was used to determine whether adult attachment style significantly predicted participants’ subjective well-being. The results of the regression analysis indicated the percentage of the variance in subjective well-being scores explained by adult attachment avoidance and adult attachment anxiety, using the following equation: y = Xβ + ε and the following model: AAttSWB.

Hypothesis 5. To test the fifth hypothesis, that perceived social support (PSS) would predict adult attachment style (AAtt), linear regression analysis was used to test if perceived social support significantly predicted participants’ adult attachment scores. The results of the regression analysis indicated the percentage of the variance in adult attachment scores explained by perceived social support, using the following equation: y = Xβ + ε and the following model:

High PSS  Low AAttAnx and High PSS  Low AAttAvo.

Hypothesis 6. To test the sixth hypothesis, that the “effects” of perceived social support

(PSS) on subjective well-being (SWB) would be partially mediated by adult attachment style

(AAtt), possible mediating effects were examined separately for each AAtt factor (anxiety and avoidance) by computing the product of the path coefficient from PSS to AAtt and the path coefficient from AAtt to SWB, using Kenny’s (2013) four-step approach to mediational analysis

109

with multiple regression. I compared the strength and statistical significance of the relationship

between each attachment score and SWB before and after PSS was included in the model to

distinguish between full and partial mediation. Full mediation would be indicated if the AAtt

style was a significant predictor of SWB when PSS was not included in the model (see test of

Hypothesis 4), but not when PSS was included in the model (Kenny, 2013). Partial mediation

would be indicated if the AAtt style was a significant predictor of SWB when PSS was excluded

from the model and was still a significant predictor, but reduced in size, when PSS was included

in the following model: (PSS  AAtt  SWB) (See Figure 9).

Hypothesis 7. To test hypothesis 7, that lower levels of perceived social support (PSS)

would strengthen the association between adult attachment (AAtt) and subjective well-being

(SWB), I used the following model: (PSS x AAtt  SWB). I chose to include a moderation

hypothesis to determine whether attachment was a mediator or a moderator of the relationship

between perceived social support and subjective well-being; the moderation analysis had the potential to lend support for the connection between perceived social support and subjective well-being. These analyses cannot be assumed to be truly causal in nature. As a result, the analyses in this study tested associations, and causal statements were then assumed to be

“causal” and effects became “effects,” acknowledging that the data collected was cross-sectional in nature as opposed to longitudinal. Instead, both the mediation and moderation analyses tested associations, instead of cause and effect.

Using Kenny’s (2013) four-step approach to moderation analysis for continuous

moderator and “causal” variables, I first determined that the causal variable (perceived social

support) was correlated and had an effect on the outcome variable (subjective well-being). This was the purpose of the third hypothesis. As previously noted with other regression analyses,

110 hierarchical linear regression was used to control for age, gender, relationship status, and education level, and each measure of subjective well-being (positive affect, negative affect, and satisfaction with life) was an outcome variable. I conducted separate hierarchical regression analyses with SWLS, PA, and NA as the criterion variables. Scores for the AAtt style and PSS were centered prior to conducting the hierarchical regression analysis to reduce multicollinearity and facilitate interpretation of any significant moderator effects (Frazier, Tix, & Barron, 2004;

Kenny, 2013). Statistically significant beta weights (p< .05) for the interaction terms and changes in R2 values with the inclusion of the interaction term were noted if present. Aiken and West’s

(1991) procedures and Kenny’s approach (2013) were used to plot and interpret the nature of any significant interaction (moderator) effects. (See Figure 10).

Hypotheses 8, 9. The procedures used to test the eighth and ninth hypotheses are articulated with Hypotheses 1 and 2.

Hypothesis 10. To test hypothesis 10, that pet attachment style (PAtt) would predict subjective well-being (SWB) after controlling for adult attachment (AAtt), I used hierarchical linear regression analysis to determine whether pet attachment scores significantly predicted participants’ subjective well-being, after controlling demographic factors and for adult attachment style scores. The results of the regression analysis indicate the percentage of the variance in subjective well-being scores explained by pet attachment avoidance and pet attachment anxiety.

Hypothesis 11. To test hypothesis 11, that perceived social support (PSS) would predict pet attachment style (PAtt, measured by pet attachment avoidance and pet attachment anxiety) after controlling for adult attachment and demographic factors, hierarchical linear regression analyses were used to test whether pet attachment scores significantly predicted participants’

111 subjective well-being. The results of the regression analysis indicated the percentage of the variance in subjective well-being scores explained by pet attachment scores.

Hypothesis 12. To test hypothesis 12, that the “effects” of perceived social support (PSS) on subjective well-being (SWB) would be partially mediated by pet attachment style (PAtt; see

Figure 11). I used Kenny’s (2013) four-step approach to determine what, if any, role that pet attachment had on the causal path between perceived social support ant subjective well-being.

Possible mediating effects were examined separately for each PAtt factor (anxiety and avoidance) by computing the product of the path coefficient from the factor to PAtt and the path coefficient from PAtt to SWB, after controlling for adult attachment (AAtt) and demographic variables. I compared the strength and statistical significance of the relationship between each attachment score and SWB before and after PSS was included in the model to distinguish between full and partial mediation. Full mediation was indicated if the PAtt style was a significant predictor of SWB when PSS was not included in the model (see test of Hypothesis

10), but not when PSS was included in the model. Partial mediation would be indicated if the

PAtt style was a significant predictor of SWB when PSS was excluded from the model and was still a significant predictor, but reduced in size, when PSS was included in the model (See Figure

11). I used the four step approach specified by Kenny because it focuses on determining whether the mediation effect is direct or indirect, and the analyses are examined in terms of zero and nonzero coefficients, as opposed to only statistical significance (Baron & Kenny, 1986). As noted by Kenny (2013), “Because trivially small coefficients can be statistically significant with large sample sizes and very large coefficients can be nonsignificant with small sample sizes, the steps should not be defined in terms of statistical significance. Statistical significance is informative, but other information should be part of statistical decision making.”

112

Hypothesis 13. To test hypothesis 13, that lower levels of perceived social support (PSS)

would strengthen the association between pet attachment (PAtt) and subjective well-being

(SWB) after controlling for adult attachment (AAtt) and demographic variables, I used the

following model: (PSS x PAtt  SWB) (See Figure 12). I chose to include a moderation

hypothesis to determine whether pet attachment was a mediator or a moderator of the

relationship between perceived social support and subjective well-being; the moderation analysis

had the potential to determine the universality of the connection between perceived social

support and subjective well-being. Using Kenny’s (2013) four-step approach to moderation

analysis for continuous moderator and causal variables, I conducted three separate hierarchical

regression analyses with SWLS, PA, and NA as the criterion variables. Scores for AAtt styles,

PAtt styles, and PSS were centered prior to conducting the hierarchical regression analysis to

reduce multicollinearity and facilitate interpretation of any significant moderator effects (Frazier,

Tix, & Barron, 2004; Kenny, 2013). Statistically significant beta weights (p< .05) for the

interaction terms and changes in R2 values when the were noted if present. Aiken and West’s

(1991) procedures and suggestions from Kenny (2013) were used to plot and interpret the nature of any significant interaction (moderator) effects (See Figure 12).

113

CHAPTER IV

RESULTS

Descriptive Statistics

Table 1 shows the means and standard deviations for each of the primary constructs in the study. The descriptive statements that follow are based on a comparison of the sample means with the maximum possible scores for each scale. Regarding adult attachment, the sample means

for the two dimensions of attachment were as follows: Adult Attachment Avoidance, 42.37 (out

of 126); Adult Attachment Anxiety, 56.83 (out of 126). On average, participants endorsed low

avoidance and anxiety in their adult romantic relationships, which suggested that they exhibited

or perceived moderate to high levels of security in their interactions and conceptualizations of

their romantic partners. The sample means for the pet attachment scales were as follows: Pet

Attachment Avoidance, 19.50 (out of 91); Pet Attachment Anxiety, 32.86 (out of 91). On

average, participants endorsed very low avoidance and low anxiety in their relationships with

their companion animals, and they exhibited high levels of secure attachment in their interactions

and conceptualizations of relationships with their selected pet. The sample mean for global

perceived social support was 50.36 (out of 60), indicating that participants reported feeling high

levels of social support from friends, family members, and significant others. The mean score for

satisfaction with life as measured by the SWLS was 26.02 (out of 35), indicating that participants

felt more satisfied than dissatisfied with their lives and the average participant was experiencing

a moderate level of satisfaction with his or her quality of life. The sample means for the PA and

NA scales were 30.49 and 14.51 (out of 50), respectively, indicating that participants reported

experiencing positive affect to a greater degree than negative affect at the time they completed

the surveys.

114

Table 1 Descriptive Statistics for Primary Constructs

Mean Standard Deviation

Adult Attachment Avoidance 42.37 19.11 Adult Attachment Anxiety 56.83 19.87 Pet Attachment Avoidance 19.50 6.59 Pet Attachment Anxiety 32.86 13.33 Global Perceived Social 50.36 6.87 Support Satisfaction With Life 26.02 5.71 Positive Affect 30.49 7.38 Negative Affect 14.51 5.01 Notes: Adult Attachment Scores are from the ECR, Pet Attachment Scores are from the PAQ, Global Perceived Social Support scores are from the MSPSS, Satisfaction with Life Scores are from the SWLS, and Affective State Scores are from the PANAS-X.

Pearson Correlations

The intercorrelations among the two adult attachment scales, two pet attachment scales, perceived social support, and the subjective well-being (SWB) variables are shown in Table 2.

Most of these correlations will be referred to later when I address the specific hypothesis analyses. The adult attachment modes were directly correlated with one another, indicating that an increase in romantic attachment avoidance coincided with anxious attachment. Adult avoidant attachment was positively correlated with both the avoidant and anxious pet attachment orientation scales. This suggests that participants’ level of attachment avoidance was positively associated with both anxious and avoidant orientations toward pets. Adult anxious attachment was also positively associated with pet anxious attachment, while the correlation between adult anxious attachment and pet avoidant attachment was not statistically significant. This indicates that participants who endorsed having an anxious attachment to their romantic partners also tended to have anxious attachments to their companion animals. Both adult attachment orientations were inversely correlated with perceived social support, satisfaction with life, and

115

positive affect and positively correlated with negative affect, indicating that as insecure (avoidant

or anxious) attachment orientations become more salient based on self-report scores, perceived social support and subjective well-being tended to decrease. This trend was also true for pet anxious attachment, in that anxious attachment to a pet was negatively correlated with indicators

of perceived social support and subjective well-being. This was not observed with pet avoidant attachment. The only significant correlation was a negative one with positive affect. Global perceived social support was inversely correlated with negative affect and positively correlated with life satisfaction and positive affect, suggesting that perceived social support is positively associated with subjective well-being. Satisfaction with life was positively correlated with

positive affect and inversely correlated with negative affect, indicating that participant’s overall

life satisfaction was related to their perceived affect. Results suggested that greater satisfaction with life was associated with more positive affect, while lower life satisfaction was correlated

with negative affect (See Table 2).

Table 2 Correlations Between Primary Constructs

Construct 1 2 3 4 5 6 7 8 1. AAttAvo 1.00

2. AAttAnx .26** 1.00

3. PAttAvo .09* .01 1.00

4. PAttAnx .18** .50** -.09** 1.00

5. Global PSS -.53** -.28** -.03 -.16** 1.00

6. SWLS -.39** -.34** .03 -.22** .44** 1.00

7. PA -.16** -.14** -.09* -.11** .17** .25** 1.00

8. NA .21** .35** .04 .32** -.20** -.36** .01 1.00

116

Notes: AAttAvo = Adult Attachment Avoidance score; AAttAnx = Adult Attachment Avoidance score; PAttAvo = Pet Attachment Avoidance score; PAttAnx = Pet Attachment Avoidance score; Global PSS= Global Perceived Social Support; SWLS = Satisfaction with Life Scale; PA= Positive Affect; NA= Negative Affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Tests of Hypotheses

Hypothesis 1. My first hypothesis had four components: lower levels of adult attachment anxiety would be associated with greater subjective well-being as represented by an inverse correlation between AAttAnx and SWLS and PA and a positive correlation between AAttAnx and NA; lower levels of adult attachment avoidance would be associated with greater subjective well-being (inverse relationship between AAttAvo and SWLS and PA and a positive correlation between AAttAvo and NA); lower levels of adult attachment anxiety would be associated with greater perceived social support (inverse correlation between AAttAnx and PSS); and lower levels of adult attachment avoidance would be associated with greater perceived social support

(inverse correlation between AAttAvo and PSS).To test hypothesis 1, that higher levels of secure adult attachment (AAtt; defined as low anxiety and low avoidance) would be positively associated with subjective well-being (SWB) and perceived social support (PSS), correlational analyses (Pearson’s r) were computed and compared.

There was some support for each component of the first hypothesis. Adult attachment anxiety was moderately negatively correlated with satisfaction with life (r= -.34, p<0.01) and negative affect (r= .35, p<0.01). The relationship between adult attachment anxiety and positive affect (r= -.14, p<0.01) was negligible but statistically significant. Similarly, adult attachment avoidance was moderately negatively correlated with satisfaction with life (r= -.39, p<0.01) and weakly positively correlated with negative affect (r= .21, p<0.01). The relationship between adult attachment avoidance and positive affect (r= -.16, p<0.01) was negligible but statistically

117 significant. Adult attachment anxiety was weakly negatively correlated with global perceived social support (r= -.28, p<0.01) and adult attachment avoidance had a strong negative correlation with global perceived social support (r= -.53, p<0.01).

Hypothesis 2. To test hypothesis 2, that global perceived social support (PSS) would be positively correlated with subjective well-being (SWB), correlational analyses (Pearson’s r) were computed. I calculated and interpreted the r’s using a one-tailed test of significance. I predicted that global perceived social support would be positively correlated with SWLS scores and PA scores and negatively correlated with NA scores. This hypothesis was supported: global perceived social support was strongly positively associated with satisfaction with life (r= .44, p<0.01), negligibly positively correlated with positive affect (r= .17, p<0.01), and weakly inversely correlated with negative affect (r= -.20, p<0.01).

The following assumptions were true with regards to the data being used in the hierarchical linear regressions for hypotheses 3, 4, and 5: all variables were quantitative; the predictor variables had a non-zero variance; there was no perfect linear relationship between two or more of the predictors; no external variables correlate with any of the variables included in the regression model; at each level of the predictor variables; the variance of the residual terms was constant (meeting the criterion definition of homoscedasticity); there was a lack of autocorrelation, meaning that residual terms were independent; the residuals in the models explored were random and normally distributed; the values of the outcome variables were independent; and the models tested were assumed to be linear (Field, 2009).

Hypothesis 3. To test hypothesis 3, that perceived social support (PSS) would be a positive predictor of subjective well-being (SWB), hierarchical linear regression analysis was used to test whether perceived social support significantly predicted participants’ subjective well-being. The

118 results of the regression analysis, shown in Tables 3 and 4, indicated the percentage of the variance in subjective well-being scores explained by perceived social support, using the following equation: y = Xβ + ε and the following model: High PSS High SWB (high SWLS, high PA, low NA). Hierarchical linear regression was used in an attempt to control for demographic variables. In the linear regression model, age, gender, relationship status, and education level were entered into the first block of the linear regression, then perceived social support was entered into the second block. Each measure of subjective well-being (positive affect, negative affect, and satisfaction with life) was an outcome variable. In the model, the equation used was Subjective Well-being (PA, NA, or SWLS)= b0 + b1 age + b2gender + b3 relstatus + b4education + b5 PSSTotal.

Table 3

Hierarchical Multiple Regression with PSS and Subjective Well-being PA NA SWLS β ΔR² β ΔR² β ΔR² Model 1 .021 .027 .027 Age .078** -.054** -.007 Gender -.772 .382 -.094 Relationship status .008 -.203 -.169 Education .027 -.324 1.030** Model 2 .038 .048 .188 Age .093** -.066** .021 Gender -.670 .304 .084 Relationship status .002 -.198 -.179 Education -.088 -.235 .828** PSS Total .211** -.162** .367** Note: PSS= Global Perceived Social Support; PA= Positive Affect; NA= Negative Affect; SWLS = Satisfaction with Life Scale *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Table 4

Model Fit for Hierarchical Multiple Regression with PSS and Subjective Well-being PA NA SWLS Model 1- Age, Gender, Relationship Status,

119

Education R .144 .164 .154 R2 .021 .027 .024 Adjusted R2 .014 .020 .016 R2 Change .021 .027 .024 F 2.839* 3.672 3.231* F Change 2.839* 3.672** 3.231 Model 2- - Age, Gender, Relationship Status, Education, PSSTotal R .242 .273 .460 R2 .058 .075 .212 Adjusted R2 .050 .066 .205 R2 Change .038 .048 .024 F 6.599** 8.570** 28.627** F Change 21.206 27.435 127.153 Durbin-Watson 1.925 1.742 .194 Note: PSS= Global Perceived Social Support; PA= Positive Affect; NA= Negative Affect; SWLS = Satisfaction with Life Scale *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

As shown in Tables 3 and 4, the linear model for the first part of the third hypothesis was

PANASpos= 19.484(Constant) + .093age - .670gender +.002relstatus -.088education

+.211PSSTotal. This model accounted for 5.8% of the variance in positive affect and was

statistically significant (p<.001). Given the Adjusted R2 of .050, the model would have a .8%

shrinkage if generalized to the population. The Durbin Watson statistic of 1.925 for the model

indicated that the assumption of independent errors was tenable. The F for the model was 6.599

(p<.001), which represented the ratio of improvement from fitting the model of perceived social

support to predict positive affect, a specific measure of subjective well-being.

The linear model for the second part of the third hypothesis was PANASneg=

25.132(Constant) -.066age +.304gender -.198relstatus -.235education -.162PSSTotal. This model accounted for 7.5% of the variance in negative affect and was statistically significant

(p<.001). Given the Adjusted R2 of .066, the model would have a .9% shrinkage if generalized to

120

the population. The Durbin Watson statistic of 1.742 for the model indicated that the assumption

of independent errors was tenable. The F for the model was 8.570 (p<.001), which represented the ratio of improvement from fitting the model of perceived social support to predict negative affect, a specific measure of subjective well-being (See Tables 3 and 4).

The linear model for the third and final part of the third hypothesis was SWLSsum=

3.604(Constant) +.021age +.084gender -.179relstatus +.828education +.367PSSTotal. This model accounted for 21.2% of the variance in satisfaction with life and was statistically significant (p<.001). Given the Adjusted R2 of .205, the model would have a .8% shrinkage if generalized to the population. The Durbin Watson statistic of 0.194 for the model indicated that the assumption of independent errors was not tenable. The F for the model was 28.627 (p<.001), which represented the ratio of improvement from fitting the model of perceived social support to predict global satisfaction with life, a specific measure of subjective well-being (See Tables 3 and 4).

Hypothesis 4. To test hypothesis 4, that adult attachment style (AAtt) would predict subjective well-being (SWB), hierarchical linear regression analysis was used to test if adult attachment style significantly predicted participants’ subjective well-being. The results of the regression analysis, as shown in Tables 5 and 6, indicated the percentage of the variance in subjective well-being scores explained by adult attachment avoidance and adult attachment anxiety. In the linear regression model, age, gender, relationship status, and education level were entered into the first block of the linear regression, then the variables for adult attachment avoidance and anxiety were entered into the second block. Measures of subjective well-being

(positive affect, negative affect, and satisfaction with life) were the outcome variables. In the

121 model, the equation used was Subjective Well-being (PA, NA, or SWLS)= b0 + b1 age + b2gender + b3 relstatus + b4education + b5 AAttAvo + b6 AAttAnx.

Table 5 Hierarchical Multiple Regression with Adult Attachment and Subjective Well-being PA NA SWLS β ΔR² β ΔR² β ΔR² Model 1 .021 .027 .024 Age .078** -.054** -.007 Gender -.772 .382 -.094 Relationship status .008 -.203 -.169 Education .027 -.324 1.030** Model 2 .038 .121 .202 Age .078** -.034* -.016 Gender -.495 .133 .374 Relationship status -.027 -.215 -.206 Education -.190 -.077 .637* AAttAvo -.064** .038** -.098** AAttAnx -.027 .073** -.070** Notes: PA= Positive Affect; NA= Negative Affect; SWLS = Satisfaction with Life Scale; AAttAvo = Adult Attachment Avoidance score; AAttAnx = Adult Attachment Avoidance score; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Table 6

Model Fit got Hierarchical Multiple Regression with AAtt and SWB PA NA SWLS Model 1- Age, Gender, Relationship Status, Education R .144 .164 .154 R2 .021 .027 .024 Adjusted R2 .014 .020 .016 R2 Change .021 .027 .024 F 2.839* 3.672 3.231* F Change 2.839* 3.672** 3.231 Model 2- - Age, Gender, Relationship Status, Education, AAttAvo, AAttAnx R .243 .385 .475 R2 .059 .148 .226 Adjusted R2 .048 .139 .217 R2 Change .038 .121 .202 F 5.548** 15.407** 25.851**

122

F Change 10.758 37.862 69.433 Durbin-Watson 1.927 1.778 .400 Notes: PA= Positive Affect; NA= Negative Affect; SWLS = Satisfaction with Life Scale; AAttAvo = Adult Attachment Avoidance score; AAttAnx = Adult Attachment Avoidance score; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Specifically, PANASpos= 34.887(Constant) + .078age - .495gender -.027relstatus -

.190education -.064AAttAvo -.027 AAttAnx. This model accounted for 5.9% of the variance in positive affect and was statistically significant (p<.001). Given the Adjusted R2 of .048, the model would have a 1.1% shrinkage if generalized to the population. The Durbin Watson statistic of 1.927 for the model indicated that the assumption of independent errors was tenable.

The F for the model was 5.548 (p<.001), which represented the ratio of improvement from fitting the model of adult attachment to predict positive affect, a specific measure of subjective well-being (See Tables 5 and 6).

In the second portion of the hypothesis test, PANASneg= 10.188(Constant) -.034age

+.133gender -.215relstatus -.077education +.038AAttAvo +.073 AAttAnx. This model accounted for 14.8% of the variance in negative affect and was statistically significant (p<.001).

Given the Adjusted R2 of .139, the model would have a .9% shrinkage if generalized to the

population. The Durbin Watson statistic of 1.778 for the model indicated that the assumption of

independent errors was tenable. The F for the model was 15.407 (p<.001), which represented the

ratio of improvement from fitting the model of adult attachment to predict negative affect, a

specific measure of subjective well-being (See Tables 5 and 6).

In the third and final portion of the fourth hypothesis, SWLSsum= 31.475(Constant) -

.016age +.374gender -.206relstatus +.637education-.098AAttAvo -.070 AAttAnx. This model

accounted for 22.6% of the variance in satisfaction with life and was statistically significant

(p<.001). Given the Adjusted R2 of .217, the model would have a .6% shrinkage if generalized to

123

the population. The Durbin Watson statistic of 0. 004 for the model indicated that the assumption

of independent errors was not tenable. The F for the model was 25.851 (p<.001), which represented the ratio of improvement from fitting the model of adult attachment to predict global satisfaction with life, a specific measure of subjective well-being (See Tables 5 and 6).

Hypothesis 5. To test hypothesis 5, that perceived social support (PSS) would predict

adult attachment style (AAtt), hierarchical linear regression analysis was used to test if perceived

social support significantly predicted participants’ adult attachment style. The results of the

regression analysis, displayed in Tables 7 and 8, indicated the percentage of the variance in

subjective well-being scores explained by adult attachment avoidance and adult attachment

anxiety. In the linear regression model, age, gender, relationship status, and education level were

entered into the first block of the linear regression, then global perceived social support

(PSSTotal) was entered into the second block. Adult attachment avoidance and anxiety were the

outcome variables. In the model, the equation used was Adult Attachment (AAttAnx or

AAttAvo)= b0 + b1 age + b2gender + b3 relstatus + b4education + b5 PSS .

Table 7

Hierarchical Multiple Regression with PSS and Adult Attachment. AAttAvo AAttAnx β ΔR² β ΔR² Model 1 .028 .062 Age .145* -.338** Gender 2.995 1.510 Relationship status -.791 .572 Education -2.530** -2.060* Model 2 .259 .102 Age .039 -.408** Gender 2.995 1.058 Relationship status -.751 .599 Education -1.748* -1.548 PSS Total -1.422** -.932**

124

Notes: AAttAvo = Adult Attachment Avoidance score; AAttAnx = Adult Attachment Avoidance score; PSS= perceived social support; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Table 8

Additional Hierarchical Multiple Regression with PSS and Adult Attachment. AAttAvo AAttAnx Model 1- Age, Gender, Relationship Status, Education R .167 .249 R2 .028 .062 Adjusted R2 .020 .055 R2 Change .028 .062 F 3.809** 8.787** F Change 3.809 8.787 Model 2- - Age, Gender, Relationship Status, Education, PSSTotal R .536 .405 R2 .287 .164 Adjusted R2 .280 .156 R2 Change .259 .102 F 42.848** 20.908** F Change 193.504 65.163 Durbin-Watson 1.969 1.845 Notes: AAttAvo = Adult Attachment Avoidance score; AAttAnx = Adult Attachment Avoidance score; PSS= perceived social support; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Specifically, AAttAvo= 116.031(Constant) + .039age +2.995gender -.751relstatus -

1.748education -1.422PSS. This model accounted for 28.7% of the variance in adult attachment avoidance and was statistically significant (p<.001). Given the Adjusted R2 of .280, the model would have a 0.7% shrinkage if generalized to the population. The Durbin Watson statistic of

1.969 for the model indicated that the assumption of independent errors was tenable. The F for the model was 42.848 (p<.001), which represented the ratio of improvement from fitting the model of perceived social support to predict adult attachment avoidance (See Tables 7 and 8).

125

In the second portion of the hypothesis test, AAttAnx= 116.589(Constant) - .408age

+1.058gender +.599relstatus -1.548education -.932PSS. This model accounted for 16.4% of the variance in adult attachment anxiety and was statistically significant (p<.001). Given the

Adjusted R2 of .156, the model would have a 0.8% shrinkage if generalized to the population.

The Durbin Watson statistic of 1.845 for the model indicated that the assumption of independent errors was tenable. The F for the model was 20.098 (p<.001), which represented the ratio of improvement from fitting the model of perceived social support to predict adult attachment anxiety (See Tables 7 and 8).

Hypothesis 6. To test hypothesis 6, that the effects of perceived social support (PSS) on subjective well-being (SWB) would be partially mediated by adult attachment style (AAtt) (PSS

 AAtt  SWB), possible mediating effects were examined separately for each AAtt factor

(anxiety and avoidance) by computing the product of the path coefficient from the factor to AAtt and the path coefficient from AAtt to SWB. I compared the strength and statistical significance of the relationship between each attachment score and SWB before and after PSS was included in the model to distinguish between full and partial mediation. Full mediation would be indicated if the AAtt style was a significant predictor of SWB when PSS was not included in the model

(see test of Hypothesis 4), but not when PSS was included in the model. Partial mediation would be indicated if the AAtt style was a significant predictor of SWB when PSS was excluded from the model and was still a significant predictor, but reduced in size, when PSS was included in the model (See Tables 7 and 8).

As explained by Kenny (2013), I followed four steps in specifying this mediational model, some of which were previous hypotheses. First, I determined that the causal variable (perceived social support) was correlated and had an effect on the outcome variable (subjective well-being).

126

This was the purpose of the third hypothesis. (See Tables 9-11 and Figure 11). The second step I took was to demonstrate that the causal variable (perceived social support) was correlated and had an effect on the hypothesized mediator (adult attachment). This was the purpose of the fifth hypothesis. As previously noted, hierarchical linear regression was used to control for age, gender, relationship status, and education level; adult attachment configurations (anxiety and avoidance) were the outcome variables. In the model, the equation used was Adult Attachment

(AAttAnx or AAttAvo)= b0 + b1 age + b2gender + b3 relstatus + b4education + b5 PSS. The linear

model accounted for 28.7% of the variance in adult attachment avoidance (p<.001) and 16.4% of

the variance in adult attachment anxiety (p<.001). The F for the models were 42.848 (p<.001)

and 20.098 (p<.001), which represented the ratio of improvement from fitting the model of

perceived social support to predict adult attachment avoidance and adult attachment anxiety,

respectively (See Tables 9-11 and Figure 11).

In the third and fourth steps of the mediational analysis I demonstrated that the mediator,

adult attachment, predicted the outcome variable, subjective well-being. In this hypothesis, it was

not sufficient to correlate the mediator (AAtt) with the outcome (SWB) because it was possible

adult attachment and subjective well-being were correlated because they were both caused by the

causal variable, perceived social support. This path was demonstrated by the fourth hypothesis.

In the third step the results of the regression analysis indicated the percentage of the variance in

subjective well-being scores explained by adult attachment avoidance and adult attachment

anxiety above and beyond perceived social support. In the linear regression model, age, gender,

relationship status, and education level were entered into the first block of the linear regression,

then total perceived social support was entered into the second block, and then the variables for

adult attachment avoidance and anxiety were entered into the third block. Measures of subjective

127

well-being (positive affect, negative affect, and satisfaction with life) were the outcome variables. In the model, as depicted in Figure 11, the equation used was Subjective Well-being

(PA, NA, or SWLS)= b0 + b1 age + b2gender + b3 relstatus + b4education + b5 PSSTotal + b6

AAttAvo + b7 AAttAnx.

Table 9

Hierarchical Multiple Regression with Perceived Social Support, Adult Attachment, and Subjective Well-being PA NA SWLS β ΔR² β ΔR² β ΔR² Model 1 .021 .027 .027 Age .078** -.054** -.007 Gender -.772 .382 -.094 Relationship status .008 -.203 -.169 Education .027 -.324 1.030** Model 2 .038 .048 .188 Age .093** -.066** .021 Gender -.670 .304 .084 Relationship status .002 -.198 -.179 Education -.088 -.235 .828** PSS Total .211** -.162** .367** Model 3 .011 .078 .066 Age .088** -.038* .001 Gender -.526 .146 .321 Relationship status -.018 -.219 -.191 Education -.189 -.078 .639* PSS Total .135* .035 .232** AAttAvo -.042* .028* -.060** AAttAnx -.018 .069** -.055** Notes: AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Table 10

Additional Hierarchical Multiple Regression with PSS, AAtt, and SWB PA NA SWLS Model 1- Age, Gender, Relationship Status, Education R .144 .164 .154

128

R2 .021 .027 .024 Adjusted R2 .014 .020 .016 R2 Change .021 .027 .024 F 2.839* 3.672 3.231* F Change 2.839* 3.672** 3.231 Model 2- - Age, Gender, Relationship Status, Education, PSSTotal R .242 .273 .460 R2 .058 .075 .212 Adjusted R2 .050 .066 .205 R2 Change .038 .048 .024 F 6.599** 8.570** 28.627** F Change 21.206 27.435 127.153 Model 3- - Age, Gender, Relationship Status, Education, PSSTotal, AAttAvo, AAttAnx R .264 .390 .528 R2 .070* .152** .278** Adjusted R2 .057 .141 .269 R2 Change .011 .078 .066 F 5.677** 13.612** 29.216** F Change 3.233 24.337 24.394 Durbin-Watson 1.942 1.785 .337 Notes: AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Specifically, PANASpos= 26.431(Constant) + .088age - .526gender -.018relstatus -

.189education + .135 PSSTotal -.042AAttAvo -.018 AAttAnx. This model accounted for 7.0% of the variance in positive affect and was statistically significant (p<.05). Given the Adjusted R2

of .057, the model would have a 1.3% shrinkage if generalized to the population. The Durbin

Watson statistic of 1.942 for the model indicated that the assumption of independent errors was tenable. The F for the model was 5.677 (p<.001), which represented the ratio of improvement

from fitting the model of adult attachment to predict positive affect, a specific measure of

129 subjective well-being. These results suggest that individuals who reported having more insecure attachment styles (higher ratings of AAttAnx and AAttAvo) had lower levels of perceived social support which was associated with decreased positive affect (lower SWB). Because the direct

“effect” of the perceived social support on positive affect remained statistically significant in the model, the relationship between perceived social support and positive affect was partially mediated by adult attachment.

In reviewing the model for negative affect, PANASneg= 13.742(Constant) -.038age

+.146gender -.219relstatus -.078education - .057PSSTotal +.028AAttAvo +.069 AAttAnx. This model accounted for 15.2% of the variance in negative affect and was statistically significant

(p<.001). Given the Adjusted R2 of .141, the model would have a 1.1% shrinkage if generalized to the population. The Durbin Watson statistic of 1.785 for the model indicated that the assumption of independent errors was tenable. The F for the model was 13.612 (p<.001), which represented the ratio of improvement from fitting the model of adult attachment to “predict” negative affect, a specific measure of subjective well-being. In other words, individuals who reported having more insecure attachment styles (higher ratings of AAttAnx and AAttAvo) had lower levels of perceived social support, which was associated with lower SWB. Because the direct “effect” of the perceived social support on negative affect became statistically non- significant in the model, the relationship between perceived social support and negative affect was fully mediated by adult attachment.

In reviewing the model for satisfaction with life, SWLSsum= 16.947(Constant) -.001age

+.321gender -.191relstatus +.639education +.232PSSTotal -.060AAttAvo -.055 AAttAnx. This model accounted for 27.8% of the variance in satisfaction with life and was statistically significant (p<.001). Given the Adjusted R2 of .269, the model would have a 0.9% shrinkage if

130

generalized to the population. The Durbin Watson statistic of 0.337 for the model indicated that

the assumption of independent errors was not tenable. The F for the model was 29.216 (p<.001),

which represented the ratio of improvement from fitting the model of adult attachment to predict

global satisfaction with life, a specific measure of subjective well-being. These results indicated

that individuals who reported having more insecure attachment styles (higher ratings of AAttAnx

and AAttAvo) had lower levels of perceived social support, which was associated with decreased

life satisfaction (lower SWB). Because the direct “effect” of perceived social support on

satisfaction with life remained statistically significant in the model, the relationship between

perceived social support and negative affect was partially mediated by adult attachment.

The sixth hypothesis was supported. Consistent with the correlations in Table 2 and the regression results in Tables 9 and 10 and as shown by the path coefficients in Figure 11, adult attachment style (AAtt, measured by adult attachment avoidance and adult attachment anxiety) partially mediated the relationship between perceived social support (PSS) and two facets of subjective well-being, positive affect and satisfaction with life. Adult attachment fully mediated the relationship between perceived social support and negative affect because the relationship between PSS and SWB became non-significant after controlling for adult attachment in the hierarchical regression. As shown in Table 11 and Figure 11, individuals who were more securely attached in their adult romantic relationships (indicated by lower scores of attachment anxiety and attachment insecurity) had higher levels of total social support, which was associated with an increased sense of self-reported subjective well-being

Table 11 Hypothesis 6, Mediational Analysis, Baron and Kenny Steps (Kenny, 2013) Step Path Estimate 95% CI β p- value 1 PSS to PA .211** .121-.300 .196 <.001 PSS to NA -.162** -.222--.101 -.221 <.001

131

PSS to SWLS .367** .303-.431 .439 <.001 2 PSS to AAttAnx -.932** -.324 <.001 PSS to AAttAvo -1.422** -.515 <.001 3 AAttAnx to PA -.018 -.052-.016 -.048 .304 AAttAnx to NA .069** .047-.091 .273 <.001 AAttAnx to SWLS -.055** -.078--.031 -.188 <.001 AAttAvo to PA -.042* -.080--.003 -.107 .034 AAttAvo to NA .028* .003-.053 .107 .026 AAttAvo to SWLS -.060** -.086--.033 -.197 <.001 4 PSS (and AAtt) to PA .135* .028-.242 .125 .014 PSS (and AAtt) to NA .057 -.126-.013 -.078 .110 PSS (and AAtt) to SWLS .232** .158-.305 .277 <.001 Notes: PSS= Perceived Social Support; AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; and NA = Negative affect. *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Figure 13.

Hypothesis 6, Mediation Diagram with Unstandardized Coefficients

AAtt -.018 AAttAnx .069**

-.055** -.932** AAttAvo -1.422 ** -.042* .028* -.060** SWB

.135* (.211**) PA .057 (-.162**) PSSTotal .232** (.367**) NA

SWLS

Notes: AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; and NA = Negative affect. *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

132

Figure 13. This model depicts the mediating role of each aspect of adult attachment

(AAttAnx and AAttAvo) in the relationship between perceived social support and each measure of subjective well-being. Note that the numbers in parentheses represent the direct effect path coefficients for the regression equation of PSSSWB, without the mediator of AAtt. These

numbers were the first step in the mediational analysis. Individuals’ attachment styles

significantly and partially mediated the relationship between perceived social support and

subjective well-being, as evidenced by the reduction in strength and significance of the

PSSSWB path coefficients after controlling for AAttAnx and AAttAvo. There was one

exception in the findings in which the path was not statistically significant, the path coefficient of

AAttAnxPA in the mediation model (β = -0.18, p= .304). Because the direct effect of PSS remained significant in two of the three measures of SWB (PA and SWLS), it indicates that attachment style partially mediated the impact of PSS on positive affect and satisfaction with life. After adding adult attachment as a mediator for the path of PSS to negative affect, the path became insignificant (p= .110), indicating that adult attachment fully mediated the relationship between PSS and negative affect. Results indicated that individuals who were more securely attached in their adult romantic relationships had higher levels of total social support, which was associated with higher levels of subjective well-being.

Hypothesis 7. To test hypothesis 7, that lower levels of perceived social support (PSS) would strengthen the association between adult attachment (AAtt) and subjective well-being

(SWB), I used the following model: (Low PSS x PAtt  SWB). I conducted three separate

hierarchical regression analyses with SWLS, PA, and NA as the criterion variables. Scores for

the AAtt style and PSS were centered prior to conducting the hierarchical regression analysis to

reduce multicollinearity and facilitate interpretation of any significant moderator effects (Frazier,

133

Tix, & Barron, 2004). I entered demographic variables in the first step; adult attachment style in

the second step; and perceived social support scores in the third step. I then entered interaction

(product) terms (AAtt style x PSS score) as the fourth step. In the model, the following equation was used: Subjective Well-being (PA, NA, or SWLS)= b0 + b1 age + b2gender + b3 relstatus +

b4education + b5AAttAvo + b6AAttAnx + b7PSSTotal + b8(PSSTotal x AAttAvo x AAttAnx).

Table 12

Hierarchical Multiple Regression with Perceived Social Support, Adult Attachment, and Subjective Well-being for Moderation, Hypothesis 7 PA NA SWLS β ΔR² β ΔR² β ΔR² Step 1 .021 .027 .024 Age .078** -.054** -.007 Gender -.772 .382 -.094 Relationship status .008 -.203 -.169 Education .027 -.324 1.030** Step 2 .038 .121 .202 Age .078** -.034* -.016 Gender -.495 .133 .374 Relationship status -.027 -.215 -.206 Education -.190 -.077 .637* AAttAvo -.064** .038** -.098** AAttAnx -.027 .073** -.070** Step 3 .011 .004 .052 Age .088** -.038* .001 Gender -.526 .146 .321 Relationship status -.018 -.219 -.191 Education -.189 -.078 .639 AAttAvo -.042* .028* -.060** AAttAnx -.027 .069** -.055** PSS Total .135* -.057 .232** Step 4 .000 .002 .002 Age .087** -.40* -.001 Gender -.549 .104 .271 Relationship status -.008 -.201 -.170 Education -.187 -.074 .644* AAttAvo -.039 .033* -.055** AAttAnx -.015 .074** -.049** PSS Total .133* -.060 .228** PSSTotal*AAttAvo*AAttAnx .000049 .000088 .000

134

Notes: AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Table 13

Additional Hierarchical Multiple Regression with PSS, AAtt, and SWB, for Moderation, Hypothesis 7 PA NA SWLS Step 1- Age, Gender, Relationship Status, Education R .144 .164 .154 R2 .021 .027 .024 Adjusted R2 .014 .020 .016 R2 Change .021 .027 .024 F 2.839* 3.672** 3.231* F Change 2.839* 3.672** 3.231* Step 2- - Age, Gender, Relationship Status, Education, AAttAvo, AAttAnx R .243 .385 .475 R2 .059 .148 .226 Adjusted R2 .048 .139 .217 R2 Change .038 .121 .202 F 5.548** 15.407** 25.851** F Change 10.758** 37.862** 69.433** Step 3- - Age, Gender, Relationship Status, Education, AAttAvo, AAttAnx, PSSTotal R .264 .390 .528 R2 .070 .152 .278 Adjusted R2 .057 .141 .269 R2 Change .011 .004 .052 F 5.677** 13.612** 29.216** F Change 6.127 2.596 38.460** Step 4- - Age, Gender, Relationship Status, Education, AAttAvo, AAttAnx, PSSTotal, (PSSTotal*AAttAvo*AAttAnx) R .265 .393 .530 R2 .070 .155 .281 Adjusted R2 .056 .142 .270 R2 Change .000 .002 .002 F 4.982 12.091** 25.828** F Change .182 1.376 1.806

135

Durbin-Watson 1.942 1.794 .343 Notes: AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

No statistically significant beta weights (p< .05) for the interaction terms or changes in R2

values from step 3 to step 4 were found. Aiken and West’s (1991) procedures would have been

used to plot and interpret the nature of any significant interaction (moderator) effects, however

there were no significant effects found for the interaction terms, which made additional

interpretation unnecessary.

As indicated in Tables 12 and 13, interaction between perceived social support and adult

attachment failed to account for a significant proportion of the variance in any of the three SWB

variables after controlling for demographic variables (age, gender, relationship status, and

educational level), the main effects of PSS, and the main effects of AAtt. This was demonstrated

by the lack of statistical significance, as well as the ΔR² values shown in Tables 12 and 13. No significant moderation effects were found in the final step of the model. The ΔR² values were low (ranging from .000 to .002) for the fourth (moderation) step, indicating that the addition of the interaction (product) term provided negligible improvement to the overall model.

The seventh hypothesis was not supported. Consistent with the regression results in

Tables 12 and 13, and as shown by the path coefficients in Figure 12, lower levels of perceived social support did not have a significant effect on the causal relationship between adult attachment style (AAtt, measured by adult attachment avoidance and adult attachment anxiety) and subjective well-being (SWB, measured by positive affect, negative affect, and satisfaction with life). These findings suggest that perceived social support does not significantly impact the relationship between pet owners’ adult attachment styles and their subjective well-being.

136

Figure 14

Moderation Model of Perceived Social Support, Adult Attachment, and Subjective Well-being- Hypothesis 7.

.133* PSSTotal -.060 .228**

SWB -.015 AAttAnx -.049** .074** PA AAttAvo -.039 .033* NA -.055**

.000049 SWLS PSSTotal .000088 x AAtt .000

Notes: AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; and NA = Negative affect.

Figure 14. This model depicts the hypothesized moderating role of the interaction between adult attachment (AAtt, measured by AAttAnx and AAttAvo) and perceived social support

(PSS) in the outcome of subjective well-being (SWB, measured by PA, NA, and SWLS). The

interaction between AAtt and PSS did not have a significant effect on SWB.

Hypothesis 8. To test hypothesis 8, that pet attachment style (PAtt) would be positively

correlated with adult attachment (AAtt) style, correlational analyses (Pearson’s r) were used. I

compared the r’s to determine whether the relationships between the following variables were

present, in the hypothesized directions, and significant at the p<.05 significance level. I predicted

that pet attachment anxiety would be positively correlated with adult attachment anxiety and that

pet attachment avoidance would be positively correlated with adult attachment avoidance. This

hypothesis was supported: pet attachment anxiety had a strong positive relationship with adult

137

attachment anxiety (r= .50, p<0.01) and pet attachment avoidance was negligibly positively

associated with adult avoidant attachment (r= .09, p<0.01). The results indicated that as pet

owners’ adult attachment styles became more anxious (and therefore less secure), their pet

attachment styles also became more anxious, and the same results were found for avoidance.

Hypothesis 9. To test hypothesis 9, that higher levels of secure pet attachment (PAtt;

defined as low anxiety and low avoidance) would be positively correlated with subjective well- being (SWB) and perceived social support (PSS), correlational analyses (Pearson’s r) were used.

I compared the r’s to determine whether the relationships between the following variables were present, in the hypothesized directions, and significant at the p<.05 significance level. I predicted that lower levels of pet attachment anxiety would be associated with greater subjective well- being [Low PAttAnx correlated with High SWB (high SWLS, high PA, low NA)]. I also hypothesized that lower levels of pet attachment avoidance will be associated with greater subjective well-being [Low PAttAvo correlated with High SWB (high SWLS, high PA, low

NA)]. A third aspect of this hypothesis was that lower levels of pet attachment anxiety would be associated with greater perceived social support (low PAttAnx correlated with High PSS). My final prediction was that lower levels of pet attachment avoidance would be associated with greater perceived social support (low PAttAvo correlated with High PSS).

There was partial support for the ninth hypothesis. Pet attachment anxiety was weakly negatively correlated with life satisfaction (r= -.22, p<0.01), negligibly related to positive affect

(r= -.11, p<0.01), and moderately positively related to negative affect (r= .32, p<0.01). The first portion of the hypothesis, that anxious attachment to a pet was associated with lower subjective well-being, was supported in that the results were statistically significant. No statistically significant correlation was found between pet attachment avoidance and life satisfaction (r= .03,

138

p=.251) or negative affect (r= .04, p=.191), but it was negligibly inversely correlated with

positive affect (r= -.09, p<0.05). Pet attachment anxiety was negligibly negatively associated

with perceived social support (r= -.16, p<0.01), indicating that as owners’ pet attachment styles

became more anxious (and therefore less secure), their sense of social support became more

diminished. The final portion of the hypothesis, that pet attachment avoidance was inversely

correlated with perceived social support, was not supported (r= -.03, p=.217), indicating no

relationship between an avoidant attachment orientation to a companion animal and an owner’s

sense of social support.

Hypothesis 10. To test hypothesis 10, that pet attachment style (PAtt) would predict

subjective well-being (SWB) after controlling for adult attachment (AAtt), hierarchical linear

regression analysis was used to test whether pet attachment scores significantly predicted

participants’ subjective well-being. In the second step of the regression, I entered the adult

attachment scores to control for their influence on subjective well-being. The results of the

regression analysis indicated the percentage of the variance in subjective well-being scores

explained by pet attachment avoidance and pet attachment anxiety. In the linear regression

model, age, gender, relationship status, and education level were entered into the first block of

the linear regression, the variables for adult attachment avoidance and anxiety were entered into

the second block, and then Pet Attachment (PAttAvo or PAttAnx) was entered into the third

block. Measures of subjective well-being (positive affect, negative affect, and satisfaction with

life) were the outcome variables. In the model, the following equation was used: Subjective

Well-being (PA, NA, or SWLS)= b0 + b1 age + b2gender + b3 relstatus + b4education + b5

AAttAvo + b6 AAttAnx + b7 PAttAvo + b8 PAttAnx.

Table 14 Hierarchical Multiple Regression with Pet Attachment and Subjective Well-being

139

PA NA SWLS β ΔR² β ΔR² β ΔR² Model 1 .021 .027 .024 Age .078** -.054** -.007 Gender -.772 .382 -.094 Relationship status .008 -.203 -.169 Education .027 -.324 1.030** Model 2 .038 .121 .202 Age .078** -.034* -.016 Gender -.495 .133 .374 Relationship status -.027 -.215 -.206 Education -.190 -.077 .637* AAttAvo -.064** .038** -.098** AAttAnx -.027 .073** -.070** Model 3 Age .071** .004 -.025 .020 -.015 .008 Gender -.550 -.018 .531 Relationship status -.037 -.181 -.216 Education -.205 -.126 .681* AAttAvo -.060** .011** -.100** AAttAnx -.024 .053** -.061** PAttAvo -.069 .016 .060* PAttAnx -.013 .064** -.025 Notes: PSS= Perceived Social Support; AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PAttAnx= Pet Attachment Anxiety Subscale; PAttAvo= Pet Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Table 15

Additional Hierarchical Multiple Regression with Pet Attachment and Subjective Well-being PA NA SWLS Model 1- Age, Gender, Relationship Status, Education R .144 .164 .154 R2 .021 .027 .024 Adjusted R2 .014 .020 .016 R2 Change .021 .027 .024 F 2.839* 3.672 3.231* F Change 2.839* 3.672** 3.231 Model 2- - Age, Gender, Relationship Status, Education, AAttAvo, AAttAnx

140

R .243 .385 .475 R2 .059 .148 .226 Adjusted R2 .028 .139 .217 R2 Change .038 .121 .202 F 5.548** 15.407** 25.851** F Change 10.758 37.862 69.433 Model 3- - Age, Gender, Relationship Status, Education, AAttAvo, AAttAnx, PAttAvo, PAttAnx R .250 .411 .484 R2 .063 .169 .234 Adjusted R2 .049 .156 .222 R2 Change .004 .020 .008 F 4.424** 13.417** 20.176** F Change 1.048 6.491 2.664 Durbin-Watson 1.909 1.799 .419

Notes: PSS= Perceived Social Support; AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PAttAnx= Pet Attachment Anxiety Subscale; PAttAvo= Pet Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Specifically, PANASpos= 36.729(Constant) + .071age - .560gender -.037relstatus -

.190education -.064AAttAvo -.027 AAttAnx -.069PAttAvo -.013PAttAnx. This model accounted for 6.3% of the variance in positive affect and was statistically significant (p<.001).

Given the Adjusted R2 of .049, the model would have a 1.4% shrinkage if generalized to the

population. The Durbin Watson statistic of 1.909 for the model indicated that the assumption of

independent errors was tenable. The F for the model was 4.424 (p<.001), which represented the

ratio of improvement from fitting the model of pet attachment to predict positive affect, a

specific measure of subjective well-being. The change in R2, which was .004, indicated that

although the addition of pet attachment scores to the model was statistically significant, it may be

less practically meaningful because the change was very small and the beta values for each pet

attachment score were non-significant. Results indicated pet attachment does not add explained

141

variance over above controls for adult attachment within its association to pet owners’ self-

reported positive affect.

In the second portion of the hypothesis test, PANASneg= 9.232(Constant) - .025age -

.018gender -.181relstatus -.126education +.034AAttAvo +.053 AAttAnx +.016PAttAvo

+.064PAttAnx. This model accounted for 16.9% of the variance in negative affect and was

statistically significant (p<.001). Given the Adjusted R2 of .156, the model would have a 1.3%

shrinkage if generalized to the population. The Durbin Watson statistic of 1.799 for the model

indicated that the assumption of independent errors was tenable. The F for the model was 13.417

(p<.001), which represented the ratio of improvement from fitting the model of pet attachment to

predict negative affect, a specific measure of subjective well-being. The R2 change statistic,

which was .020, indicated that although the addition of pet attachment scores to the model was

statistically significant, it may be less practically meaningful because the change was very small

and the beta value for pet attachment avoidance was non-significant. In the model, pet attachment anxiety (PAttAnx) did have a statistically significant association on negative affect (β

= 0.064, p<.01), indicating that anxious pet attachment was a positive predictor of negative affect and mood.

In the third and final portion of the tenth hypothesis, SWLSsum= 30.180(Constant) -

.015age +.531gender -.216relstatus +.681education -.100AAttAvo -.061 AAttAnx

+.060PAttAvo -.025PAttAnx. This model accounted for 23.4% of the variance in satisfaction

with life and was statistically significant (p<.001). Given the Adjusted R2 of .222, the model

would have a 1.2% shrinkage if generalized to the population. The Durbin Watson statistic of

0.419 for the model indicated that the assumption of independent errors was not tenable. The F

for the model was 20.176 (p<.001), which represented the ratio of improvement from fitting the

142

model of adult attachment to predict global satisfaction with life, a specific measure of subjective

well-being. The R2 change statistic, which was .008, indicated that although the addition of pet

attachment scores to the model was statistically significant, it may be less practically meaningful

because the change was very small and the beta value for the pet attachment anxiety score was

non-significant. In the model, pet attachment avoidance (PAttAvo) did have a statistically

significant impact on satisfaction with life (β = 0.060, p<.05), indicating that avoidant pet

attachment was a positive predictor of overall satisfaction with life. This finding was in contrast

with the tenth hypothesis, which held that secure pet attachment, and therefore low pet

attachment avoidance, would predict greater satisfaction with life.

Overall, two of the six portions of the tenth hypothesis were supported. Pet attachment

anxiety did not significantly predict positive affect or satisfaction with life, though the models

were statistically significant. Pet attachment avoidance did not significantly predict positive or

negative affect, although the models were statistically significant. Pet attachment anxiety did

significantly predict negative affect, and this was in the hypothesized direction. Pet attachment

avoidance significantly predicted satisfaction with life, but this was not in the hypothesized

direction. And overall, although the regression models were significant, the improvement of the

model was negligible compared to not using pet attachment scores.

Hypothesis 11. To test hypothesis 11, that perceived social support (PSS) would predict

pet attachment style (PAtt) after controlling for adult attachment, hierarchical linear regression

analyses were used to test whether PSS scores significantly predicted participants’ PAttAvo and

PAttAnx scores. In the second step of the regression, I entered the adult attachment scores to control for their influence on pet attachment scores. The results of the regression analysis indicated the percentage of the variance in pet attachment scores explained by perceived social

143

support scores. In the linear regression model, age, gender, relationship status, and education

level were entered into the first block of the linear regression, the variables for adult attachment

avoidance and anxiety were entered into the second block, and then perceived social support

(PSSTotal) was entered into the third block. Pet attachment scores (PAttAvo and PAttAnx) were the outcome variables. In the model, the equation used was Pet Attachment (PAttAnx or

PAttAvo)= b0 + b1 age + b2gender + b3 relstatus + b4education + b5 AAttAvo + b6 AAttAnx +

b7PSS.

Table 16

Hierarchical Multiple Regression with PSS and Pet Attachment. PAttAvo PAttAnx β ΔR² β ΔR² Model 1 .026 .064 Age -.064** -.233** Gender -1.332 3.353* Relationship status -.099 -.373 Education -.422 .098 Model 2 .017 .224 Age -.078** -.133** Gender -1.471* 2.720 Relationship status -.049 -.520 Education -.370 .855 AAttAvo .047** .043 AAttAnx -.023 .315** Model 3 Age -.079** .000 -.134** .000 Gender -1.470* 2.722* Relationship status -.049 -.520 Education -.371 .855 AAttAvo .046** .041 AAttAnx -.023 .314** PSSTotal -.003 -.011

Notes: PSS= Perceived Social Support; AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PAttAnx= Pet Attachment Anxiety Subscale; PAttAvo= Pet Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

144

Table 17

Additional Hierarchical Multiple Regression with PSS and Pet Attachment. PAttAvo PAttAvo Model 1- Age, Gender, Relationship Status, Education R .160 .253 R2 .026 .064 Adjusted R2 .018 .057 R2 Change .026 .064 F 3.490** 9.092** F Change 3.490 9.092 Model 2- - Age, Gender, Relationship Status, Education, AAttAvo, AAttAnx R .206 .537 R2 .042 .288 Adjusted R2 .032 .280 R2 Change .017 .224 F 3.928** 35.782** F Change 4.707 85.532 Model 3- - Age, Gender, Relationship Status, Education, AAttAvo, AAttAnx, PSSTotal R .206 .537 R2 .043 .288 Adjusted R2 .030 .279 R2 Change .000 .000 F 3.361** 30.616** F Change .004 .015 Durbin-Watson 2.037 1.873

Notes: PSS= Perceived Social Support; AAttAnx= Adult Attachment Anxiety Subscale; AAttAvo= Adult Attachment Avoidance Subscale; PAttAnx= Pet Attachment Anxiety Subscale; PAttAvo= Pet Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Specifically, PAttAvo = 25.296 -.079age -1.470gender -.049relstatus -.371education

+.046AAttAvo -.023AAttAnx -.003PSSTotal. This model accounted for 4.3% of the variance in pet attachment avoidance and was statistically significant (p<.001). Given the Adjusted R2 of

.030, the model would have a 1.3% shrinkage if generalized to the population. The Durbin

145

Watson statistic of 2.037 for the model indicated that the assumption of independent errors was

tenable. The F for the model was 3.361 (p<.01), which represented the ratio of improvement

from fitting the model of perceived social support to predict adult attachment avoidance. The R2

change statistic, which was .000, was negligible, suggesting that although the model was

statistically significant, it may be less practically meaningful because the change was very small and the overall model only accounted for 4.3% of the total variance in pet attachment avoidance.

The addition of perceived social support to the model had minimal impact on pet attachment

avoidance scores, and PSS did not have a statistically significant role in the regression equation

(β = -0.003, p=0.951).

In the second portion of the hypothesis test, PAttAnx = 9.186 -.134 age + 2.722gender -

.250relstatus + .855education + .041AAttAvo + .314AAttAnx -.011PSS. This model accounted

for 28.8% of the variance in pet attachment anxiety and was statistically significant (p<.001).

Given the Adjusted R2 of .279, the model would have a 0.9% shrinkage if generalized to the

population. The Durbin Watson statistic of 1.873 for the model indicated that the assumption of

independent errors was tenable. The F for the model was 30.616 (p<.001), which represented the

ratio of improvement from fitting the model of perceived social support to predict adult

attachment anxiety. The R2 change statistic, which was .000, was negligible, suggesting that

although the overall model was statistically significant, it may have little practical meaning

because the change was very small and the overall model accounted for 28.8% of the total

variance in pet attachment anxiety. The addition of perceived social support to the model had

minimal impact on pet attachment anxiety scores, because both Model 2 (without PSS) and

Model 3 (with PSS) were nearly identical, and PSS did not have a statistically significant role in

the regression equation (β = -0.011, p=0.901).

146

Hypothesis 11 was not supported. The addition of perceived social support to the regression model did not significantly predict pet attachment style, despite the overall models being significant. In sum, perceived social support did not significantly predict participants’ pet attachment styles.

Hypothesis 12. To test hypothesis 12, that the effects of perceived social support (PSS) on subjective well-being (SWB) would be partially mediated by pet attachment style (PAtt) (PSS 

PAtt  SWB) above and beyond adult attachment, possible mediating effects were examined separately for each PAtt factor (anxiety and avoidance) by computing the product of the path coefficient from the factor to PAtt and the path coefficient from PAtt to SWB after controlling for adult attachment and demographic variables. I compared the strength and statistical significance of the relationship between each attachment score and SWB before and after PSS was included in the model to distinguish between full and partial mediation. Full mediation was indicated if the PAtt style was a significant predictor of SWB when PSS was not included in the model (see test of Hypothesis 10), but not when PSS was included in the model. Partial mediation would be indicated if the PAtt style was a significant predictor of SWB when PSS was excluded from the model and was still a significant predictor, but reduced in size, when PSS was included in the model.

As explained by Kenny (2013) and as completed in the sixth hypothesis, I followed four steps in specifying this mediational model, most of which were previous hypotheses. First, I determined that the causal variable (perceived social support) was correlated and had an effect on the outcome variable (subjective well-being). This was the purpose of the third hypothesis. As previously noted, hierarchical linear regression was used to control for the demographic variables of age, gender, relationship status, and education level, and each measure of subjective

147

well-being (positive affect, negative affect, and satisfaction with life) was an outcome variable.

In the model, the equation used was Subjective Well-being (PA, NA, or SWLS)= b0 + b1 age +

b2gender + b3 relstatus + b4education + b5 AAttAvo + b6 AAttAnx + b7 PAttAvo + b8 PAttAnx.

The linear model accounted for 5.8% of the variance in positive affect (p<.001), 7.5% of the

variance in negative affect (p<.001), and 21.2% of the variance in satisfaction with life (p<.001).

The F for the models were 6.599 (p<.001), 8.570 (p<.001), and 28.627 (p<.001), which

represented the ratio of improvement from fitting the model of perceived social support to

predict positive affect, negative affect, and satisfaction with life, respectively.

The second step I took was to demonstrate that the causal variable (perceived social

support) was correlated and had an effect on the hypothesized mediator (pet attachment), above

and beyond adult attachment. This was the purpose of the eleventh hypothesis, and results of this

analysis were mixed. As previously noted, hierarchical linear regression was used to control for

age, gender, relationship status, and education level; adult attachment configurations (anxiety

and avoidance) were the outcome variables. In the model, the equation used was Pet Attachment

(PAttAnx or PAttAvo)= b0 + b1 age + b2gender + b3 relstatus + b4education + b5 AAttAvo + b6

AAttAnx + b7PSS. The linear model accounted for 4.3% of the variance in pet attachment

avoidance (p<.001) and 28.8% of the variance in pet attachment anxiety (p<.001). The F for the models were 3.361 (p<.01) and 30.616 (p<.001), which represented the ratio of improvement

from fitting the model of perceived social support to predict pet attachment avoidance and pet

attachment anxiety, respectively. Though the overall models were statistically significant, it is

also important to consider the role of perceived social support in each outcome of pet attachment.

In particular, the ΔR² was .000 for both pet attachment anxiety and pet attachment avoidance,

meaning that adding perceived social support to the model explained no additional variance in

148

pet attachment orientation (See Tables 16 and 17). In contrast to what was expected, perceived

social support was not a predictor of pet attachment, meaning that an individual’s sense of

perceived social support had no significant relationship on his or her pet attachment. Results

indicated pet attachment does not add explained variance over above controls for adult

attachment within its association to pet owners’ subjective well-being.

Due to the fact that the second step of the mediation analysis was unable to be completed, it was no longer meaningful or feasible to complete the remaining portions of the twelfth hypothesis. I was not able to complete the third or fourth steps of the mediational analysis, in which I would have demonstrated that the mediator, pet attachment (PAtt), affected the outcome variable, subjective well-being. Hypothesis 12 was not supported. As demonstrated by the fourth hypothesis, and as depicted in Figure 13, the first step of the mediation analysis, which showed that PSS was a significant predictor of each facet of SWB, was successful. The second step of the mediation (which was the tenth hypothesis), which demonstrated that PSS had a significant effect on PAtt, was not supported. Because this threshold was not met, the final two steps of mediation were not completed.

149

Figure 15.

Pet Attachment Mediation Diagram with Unstandardized Coefficients

PAtt (-.013) PAttAnx (.064**)

(-.025) -.011 PAttAvo -.003 (-.069) (.016) (.060*) SWB

.211** PA -.162** PSSTotal .367** NA

SWLS

Notes: PAttAnx= Pet Attachment Anxiety Subscale; PAttAvo= Pet Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; and NA = Negative affect. *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Figure 15. This model depicts the hypothesized mediating role of each aspect of pet

attachment (PAttAnx and PAttAvo) in the relationship between perceived social support and

each measure of subjective well-being. Note that the direct effect path coefficients for the

regression equation of PSSSWB were the first step in the mediational analysis; these numbers

were statistically significant for each measure of SWB. The second step of the mediational

analysis involved computing the direct effect path coefficients for the regression equation of

PSSPAtt. Neither of these paths was statistically significant, suggesting that the model was

incorrect. The numbers in parentheses, which represent the direct effect of each pet attachment

style on SWB measures (without including PSS), were the results of the eleventh hypothesis.

150

Hypothesis 13. To test hypothesis 13, that lower levels of perceived social support (PSS)

would strengthen the association between pet attachment (PAtt) and subjective well-being

(SWB) after controlling for adult attachment (AAtt), I conducted three separate hierarchical

regression analyses with SWLS, PA, and NA as the criterion variables. I hypothesized that the

relationship between PAtt and SWB would be stronger for participants who reported lower

ratings of PSS. Scores for AAtt styles, PAtt styles, and PSS were centered prior to conducting the hierarchical regression analysis to reduce multicollinearity and facilitate interpretation of any significant moderator effects (Frazier, Tix, & Barron, 2004). I also controlled for demographic variables (age, gender, relationship status, and educational attainment) in the first step; adult attachment style in the second step; pet attachment style in the third step; and perceived social support scores in the fourth step. I then investigated interaction (product) terms (i.e PAtt style x

PSS score) as the fifth step. Statistically significant beta weights (p< .05) for the interaction

terms and changes in R2 values from step 4 to step 5 were noted if present. In the model, the

following equation was used: Subjective Well-being (PA, NA, or SWLS)= b0 + b1 age +

b2gender + b3 relstatus + b4education + b5AAttAvo + b6AAttAnx + b7PAttAvo + b8PAttAnx +

b9PSSTotal + b10(PSSTotal x PAttAvo x PAttAnx).

Table 18

Hierarchical Multiple Regression with Perceived Social Support, Pet Attachment, and Subjective Well-being for Moderation, Hypothesis 13 PA NA SWLS β ΔR² β ΔR² β ΔR² Model 1 .021 .027 .024 Age .078** -.054** -.007 Gender -.772 .382 -.094 Relationship status .008 -.203 -.169 Education .027 -.324 1.030** Model 2 .038 .121 .202 Age .078** -.034* -.016 Gender -.495 .133 .374

151

Relationship status -.027 -.215 -.206 Education -.190 -.077 .637* Ctr AAttAvo -.064** .038** -.098** Ctr AAttAnx -.027 .073** -.070** Model 3 .004 .020 .008 Age .071** -.025 -.015 Gender -.560 .018 .531 Relationship status -.037 -.181 -.216 Education -.205 -.126 .681** Ctr AAttAvo -.060** .034** -.100** Ctr AAttAnx -.024 .053** -.061** Ctr PAttAvo -.069 .016 .060 Ctr PAttAnx -.013 .064** -.025 Model 4 .011 .004 .052 Age .080** -.029 .002 Gender -.592 -.004 .477 Relationship status -.028 -.185 -.201 Education -.203 -.127 .683** Ctr AAttAvo -.038 .025* -.062** Ctr AAttAnx -.015 .050** -.045** Ctr PAttAvo -.069 .016 .061 Ctr PAttAnx -.013 .064** -.024 Ctr PSS Total .134* -.056 .231** Model 5 .000 .004 .003 Age .081** -.028 .002 Gender -.585 .020 .451 Relationship status -.027 -.182 -.204 Education -.204 -.131 .687** Ctr AAttAvo -.038 .025* -.062** Ctr AAttAnx -.015 .050** -.046** Ctr PAttAvo -.072 .005 .072* Ctr PAttAnx -.013 .063** -.023 Ctr PSS Total .135* -.056 .231* PSSTotal*PAttAvo*PAttAnx .000 -.001 .001 Notes: Ctr AAttAvo= Adult Attachment Avoidance Subscale (centered); Ctr AAttAnx= Adult Attachment Anxiety Subscale (centered); Ctr PAttAvo= Pet Attachment Avoidance Subscale (centered); Ctr PAttAnx= Pet Attachment Anxiety Subscale (centered); Ctr PSSTotal= Total Perceived Social Support (centered); SWLS = Satisfaction with Life Scale; PA = Positive affect; NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

Table 19

Hierarchical Multiple Regression with Perceived Social Support, Pet Attachment, and Subjective Well-being, for Moderation, Hypothesis 13 PA NA SWLS

152

Model 1- Age, Gender, Relationship Status, Education R .144 .164 .154 R2 .021 .027 .024 Adjusted R2 .014 .020 .016 R2 Change .021 .027 .024 F 2.839* 3.672** 3.231* F Change 2.839* 3.672** 3.231* Model 2- - Age, Gender, Relationship Status, Education, Ctr AAttAvo, Ctr AAttAnx R .243 .385 .475 R2 .059 .148 .226 Adjusted R2 .048 .139 .217 R2 Change .038 .121 .202 F 5.548** 15.407** 25.851** F Change 10.758** 37.862** 69.433** Model 3- - Age, Gender, Relationship Status, Education, Ctr AAttAvo, Ctr AAttAnx, Ctr PAttAnx, Ctr PAttAvo R .250 .411 .484 R2 .063 .169 .234 Adjusted R2 .049 .156 .222 R2 Change .004 .020 .008 F 4.424** 13.417** 20.176** F Change 1.048 6.491** 2.264 Model 4- - Age, Gender, Relationship Status, Education, AAttAvo, AAttAnx, Ctr PAttAnx, Ctr PAttAvo, PSSTotal R .271 .416 .535 R2 .073 .173 .286 Adjusted R2 .058 .159 .274 R2 Change .011 .004 .052 F 4.647** 12.245** 23.512** F Change 6.096 2.555 38.699** Model 5- - Age, Gender, Relationship Status, Education, AAttAvo, AAttAnx, Ctr PAttAnx, Ctr PAttAvo, PSSTotal, (PSSTotal x AAttAvo x AAttAnx)

153

R .271 .420 .538 R2 .074 .177 .289 Adjusted R2 .056 .161 .276 R2 Change .000 .004 .003 F 4.184** 11.313** 21.469** F Change .082 2.589 2.486 Durbin-Watson 1.922 1.805 .372 Notes: Ctr AAttAvo= Adult Attachment Avoidance Subscale (centered); Ctr AAttAnx= Adult Attachment Anxiety Subscale (centered); Ctr PAttAvo= Pet Attachment Avoidance Subscale (centered); Ctr PAttAnx= Pet Attachment Anxiety Subscale (centered); Ctr PSSTotal= Total Perceived Social Support (centered); SWLS = Satisfaction with Life Scale; PA = Positive affect; NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level.

As displayed in Tables 18 and 19, no statistically significant beta weights (p< .05) for the interaction terms or changes in R2 values from step 4 to step 5 were found. Aiken and West’s

(1991) procedures would have been used to plot and interpret the nature of any significant interaction (moderator) effects, however there were no significant effects found for the interaction terms, which made additional interpretation unnecessary.

As indicated in Tables 18 and 19, interaction between perceived social support and pet attachment failed to account for a significant proportion of the variance in any of the three SWB variables after controlling for demographic variables (age, gender, relationship status, and educational level), the main effects of PSS, and the main effects of PAtt, above and beyond adult attachment. This was demonstrated by the lack of statistical significance, as well as the ΔR² values shown in Tables 18 and 19. No significant moderation effects were found in the final step of the model. The ΔR² values were low (ranging from .000 to .004) for the fifth (moderation) step, indicating that the addition of the interaction (product) term provided negligible improvement to the overall model.

The thirteenth hypothesis was not supported. Consistent with the regression results in

Tables 18 and 19, and as shown by the path coefficients in Figure 14, lower levels of perceived

154

social support did not have a significant effect on the causal relationship between pet attachment

style (PAtt, measured by pet attachment avoidance and pet attachment anxiety) and subjective

well-being (SWB, measured by positive affect, negative affect, and satisfaction with life). These

findings suggest that perceived social support does not significantly impact the relationship

between pet owners’ pet attachment styles and their subjective well-being.

Figure 16

Moderation Model of Perceived Social Support, Pet Attachment, and Subjective Well-being- Hypothesis 13.

.135* PSSTotal -.056 .231**

-.013 SWB PAttAnx .074** -.023 PA PAttAvo -.072 .005 NA .072*

.000 SWLS PSSTotal -.001 x PAtt .001

Notes: PAttAnx= Pet Attachment Anxiety Subscale; PAttAvo= Pet Attachment Avoidance Subscale; PSSTotal= Total Perceived Social Support; SWLS = Satisfaction with Life Scale; PA = Positive affect; and NA = Negative affect; *denotes statistical significance at .05 level; ** denotes statistical significance at .01 level. Figure 16. This model depicts the hypothesized moderating role of the interaction

between pet attachment (PAtt, measured by PAttAnx and PAttAvo) and perceived social support

(PSS) in the outcome of subjective well-being (SWB, measured by PA, NA, and SWLS). The

interaction between PAtt and PSS did not have a significant effect on SWB.

155

CHAPTER 5

DISCUSSION

The results of the present study provided empirical evidence for the complex

relationships of subjective well-being, perceived social support, adult attachment, and

companion animal attachment. With some exceptions, the findings were consistent with past

research that demonstrated a link between adult attachment, subjective well-being, and perceived

social support. The findings also provided exploratory confirmation of theoretical research which

identified linkages between adult attachment, pet attachment, subjective well-being, and

perceived social support.

Summary and Interpretation of Results

Hypothesis 1. In the first hypothesis, I predicted that lower levels of adult attachment

anxiety would be associated with greater subjective well-being, lower levels of adult attachment avoidance would be associated with greater subjective well-being, lower levels of adult attachment anxiety would be associated with greater perceived social support, and lower levels of adult attachment avoidance would be associated with greater perceived social support. All sub-components of this hypothesis were supported and statistically significant. Adult attachment avoidance was significantly positively correlated with negative affect and was inversely correlated with perceived social support, satisfaction with life, and positive affect. Adult attachment anxiety was significantly positively correlated with negative affect and was inversely correlated with perceived social support, satisfaction with life, and positive affect. The strength of the relationships varied from negligible to strong.

The analyses of the first hypothesis confirm and support previously published research which suggests that adult attachment relationships can provide extensive social support

156

(Mikulincer & Shaver, 2007) and are linked with well-being (Sadava et al., 2009). Perceived

social support from a romantic partner has been consistently associated with an increase in one’s

perceived ability to cope with stress, reducing anxiety and rumination, and a significant

reduction in one’s experience of pain (Cohen & McKay, 1984; MacDonald & Leary, 2005;

Thoits, 1986). Within the realm of secure attachment orientations, researchers have found

consistent links between relationship security, perceived social support, and elevated mood

(enhanced well-being). Results from the current study indicate that secure adult attachment

orientations, as defined by lower levels of anxious attachment and lower levels of avoidant

attachment, are associated with high levels of perceived social support and an increased sense of

subjective well-being in pet-owning adults. These correlations ranged from negligible to strong,

and all were statistically significant. The results suggest that adult attachment style is

meaningfully related to satisfaction with life and with perceived social support, which could have

important implications in counseling psychology. Although the relationship is known,

directionality and clinical implications cannot be inferred from correlational analysis alone.

Perceived social support, attachment style, and subjective well-being are constructs which are often assessed by counselors. Knowledge that they are related could be helpful in constructing effective interventions for clients or when conceptualizing client concerns. Findings from the current study suggest that the sample used was similar in response style and patterns to previous research (Mikulincer & Shaver, 2007; Sadava et al., 2009).

Hypothesis 2. In Hypothesis 2, I expected a positive correlation between perceived social support and subjective well-being among survey participants. In the current study, this prediction was supported for all three measures of subjective well-being. Perceived social support was significantly inversely correlated with negative affect (r= -.20, p<0.01) and it was significantly

157

positively correlated with positive affect (r= .17, p<0.01) and satisfaction with life (r= .44,

p<0.01). The strength of the relationships varied from negligible to moderate.

The results of the second hypothesis analysis support and confirm the research of Cohen

and Wills (1985), whose literature review suggested that the presence and availability of a

support system reduced the negative impact of distress for adults. Their research also indicated

that when participants believed they were a part of a social network, they reported an enhanced

sense of subjective well-being. The results of the second hypothesis are both practically and

statistically significant. Although causality cannot be determined through the correlational

analyses, the strength of the association between perceived social support and subjective well-

being suggests that the constructs are both independent (because r<1.00) and connected to one

another (because r>0). This information could be helpful for counseling psychologists in

conceptualizing their clients and in assessing multiple domains of functioning. It could also be

beneficial to consider the importance of assessing individuals’ senses of perceived social

support, given its relationship to self-reported subjective well-being.

Hypothesis 3. In the third hypothesis, I expected perceived social support to be a positive

predictor of subjective well-being among the study participants. The hypothesis was supported

for all three measures of subjective well-being. As with the first two hypotheses, this was confirmatory data analysis to determine whether the study’s sample was similar in response

pattern and style to other samples in the literature (Cohen & McKay, 1984; Deci et al., 2006;

MacDonald & Leary, 2005; Sadava et al., 2009; Thoits, 1986). All components of the third

hypothesis were supported, even after controlling for the demographic variables of age, gender,

relationship status, and education level. Perceived social support accounted for 5.8% of the

variance in positive affect, 7.5% of the variance in negative affect, and 21.2% of the variance in

158

satisfaction with life, and all were statistically significant at the p<0.001 level. Additionally, the

F for each of the models ranged from 6.599 to 28.627 and were also statistically significant at the

p<0.01 levels. As expected, perceived social support was a positive predictor of both satisfaction

with life and positive affect, and a negative predictor of negative affect, meaning that adults who

expressed having higher levels of perceived social support subsequently reported enhanced

subjective well-being. These findings indicated that the perceived social support has a significant

effect on subjective well-being, and it does so in ways which are both statistically and practically

meaningful (See Tables 3 and 4). The results of the third hypothesis analysis support and confirm

the research of Cohen and McKay (1984), Deci and colleagues (2006), MacDonald and Leary

(2005, Sadava and colleagues (2009), and Thoits (1986). Such findings could have important

practical implications. Regardless of age, gender, relationship status, or educational background,

it may be possible to positively impact an adult’s sense of subjective well-being by enhancing

that individual’s sense of social support. In clinical populations, therapists often gauge progress

in therapy by assessing individuals’ overall functioning and sense of mental health. One important aspect of mental health is the individual’s own appraisal of his or her subjective well-

being. By increasing a person’s sense of perceived social support, or their sense of comfort and

nurturing from friends, family members, and significant others, that individual could experience

a significant decrease in negative affect or an increase in his or her positive affect or satisfaction

with life.

Hypothesis 4. In the fourth hypothesis, I expected adult attachment style to be a positive predictor of subjective well-being among the study participants. The hypothesis was supported in five of the six paths tested. Adult attachment avoidance significantly predicted all three measures

of subjective well-being, and adult attachment anxiety avoidance significantly predicted two of

159

the three measures of subjective well-being, negative affect and satisfaction with life. When combined in the regression models, overall adult attachment style significantly predicted all three subjective well-being outcomes.

Most components of the fourth hypothesis were supported, even after controlling for the demographic variables of age, gender, relationship status, and education level. Adult attachment

(combining both anxiety and avoidance) accounted for 5.9% of the variance in positive affect,

14.8% of the variance in negative affect, and 22.6% of the variance in satisfaction with life, and all were statistically significant at the p<0.001 level. As expected, adult attachment was a positive predictor of both satisfaction with life and positive affect, and a negative predictor of negative affect, meaning that adults who expressed having more secure adult attachment styles subsequently reported enhanced subjective well-being. These findings indicated that adult attachment style predicts higher ratings of subjective well-being, and it does so in ways which are both statistically and practically meaningful (See Tables 5 and 6). The results of the fourth hypothesis analysis support and confirm previous research on health outcomes and adult attachment style (Coan, 2010; Sadava et al., 2009; Siegel, 2001; Strathearn, 2011). Such findings could have important practical implications. Regardless of age, gender, relationship status, or educational background, it may be possible to positively impact an adult’s sense of subjective well-being by decreasing that individual’s attachment anxiety and/or attachment avoidance. In therapy, relationships with others are often a focus area and these results suggest that it may be possible to improve an adult’s sense of subjective well-being by increasing the strength of secure attachment bonds in adult relationships. In other words, by increasing a client’s sense of attachment to others, or by enhancing the sense that important romantic relationships can provide

160 a secure base or safe haven, for example, the client may then experience a significant decrease in negative affect or an increase in his or her positive affect or satisfaction with life.

Hypothesis 5. In the fifth hypothesis, I anticipated that perceived social support would be a positive predictor of adult attachment style among the study participants. The hypothesis was supported. After controlling for the demographic variables of age, gender, relationship status, and education level, perceived social support accounted for 28.7% of the variance in adult attachment avoidance and 16.4% of the variance in adult attachment anxiety, and both were statistically significant at the p<0.001 level. As expected, perceived social support was a negative predictor of adult attachment avoidance and anxiety, meaning that adults who endorsed having more social support from friends, families, and partners subsequently reported secure adult attachment styles (defined by lower anxiety and avoidance in attachment orientation). These findings indicated that perceived social support has a significant effect on adult attachment style, and it does so in ways which are both statistically and practically important (See Tables 7 and 8).

The results of the fifth hypothesis analysis expand previous research on attachment style

(Mikulincer & Shaver, 2007) and social support (Sadava et al., 2009). Such conclusions could have important practical applications. Regardless of age, gender, relationship status, or educational background, it may be possible to increase the security and health of an adult’s attachment style by increasing that individual’s perceptions of available social support from friends, family members, and significant others. In therapy, relationships with others are often a focus area and these results suggest that it may be possible to increase the strength, stability, and security of an adult’s attachment bonds in adult relationships by enhancing his or her sense of social support. In other words, by increasing someone’s belief that he or she has a support system which can be reliably accessed, the individual may then experience an enhanced sense that

161 important romantic relationships can provide a secure base or safe haven.

Hypothesis 6. In the sixth hypothesis, I predicted that the effects of perceived social support on subjective well-being would be partially mediated by adult attachment style. After following Kenny’s (2013) four steps for mediation, and in controlling for the demographic variables of age, gender, relationship status, and education level, I demonstrated that individuals who reported having more insecure attachment styles (higher ratings of attachment avoidance and anxiety) had lower levels of perceived social support, which was associated with lower levels of subjective well-being (decreased positive affect, increased negative affect, and lower overall life satisfaction). As noted in the analyses for Hypothesis 6 and Tables 9-11, adult attachment partially mediated the relationship between perceived social support and positive affect, fully mediated the relationship between perceived social support and negative affect, and partially mediated the relationship between perceived social support and overall satisfaction with life. The results suggest that adults who had less social support had lower levels of subjective well-being, and they had less secure attachment orientations. Perhaps these individuals’ sense of support from friends, family, and significant others was impacted by avoidance or anxiety in their attachment orientations. The fact that attachment style fully mediated the relationship between social support and negative affect suggests that a person’s attachment style may have a significant influence on that person’s perception of negative emotionality beyond the sense of acceptance felt from the support group or system.

In the empirical research that examines the mediating role of adult attachment style on the relationship between perceived social support and subjective well-being, these results are consistent with findings that attachment orientation has a direct connection to physical well- being (Sadava et al., 2009) and that the physical presence of a close friend or romantic partner

162

during a stressful experience is associated with a reduction in stress response (Coan, Schaefer, &

Davidson, 2006; Edens, Larkin, & Abel, 1992; Kamarck, Manuck, & Jennings, 1990). Within

the realm of secure attachment orientations, researchers have found consistent links between

relationship security, perceived social support, and elevated mood (enhanced well-being). In adults, even the symbolic availability of attachment figures, through priming of attachment figure representations or asking a participant to visualize an attachment figure, has a soothing effect (Mikulincer & Shaver, 2007). As previously noted, other researchers have found evidence that negative or distressing interactions with romantic partners negatively impacted subjective well-being. These unpleasant interactions could be perceived criticism or rejection from a romantic partner, and effects included increased levels of depressed mood, anger, and anxiety in participants (Finch et al., 1999). The results of the sixth hypothesis support these findings by suggesting that that individuals who reported having more insecure attachment styles had lower levels of perceived social support, which, in turn, led to lower levels of subjective well-being.

Hypothesis 7. In the seventh hypothesis, I predicted that lower levels of perceived social support would strengthen (moderate) the association between adult attachment and subjective well-being. This hypothesis was not supported, because no significant moderation effects or changes in R2 were found for the interaction between adult attachment and perceived social

support in predicting participants’ subjective well-being scores. Consistent with the regression

results in Tables 12 and 13, and as shown by the path coefficients in Figure 12, lower levels of

perceived social support did not have a significant effect on the causal relationship between adult

attachment style (AAtt, measured by adult attachment avoidance and adult attachment anxiety)

and subjective well-being. These findings suggest that perceived social support does not

significantly impact the relationship between pet owners’ adult attachment styles and their

163

subjective well-being. The results of the current study suggest that mediation, as shown in the

sixth hypothesis, may be the correct model for understanding the complex relationship between

adult attachment, perceived social support, and subjective well-being, rather than moderation, as

shown in the seventh hypothesis.

Hypothesis 8. In the eighth hypothesis, I anticipated that pet attachment style would be

positively correlated with adult attachment style. This hypothesis was somewhat supported: pet

attachment anxiety had a strong positive relationship with adult attachment anxiety (r= .50,

p<0.01) and a negligible but significant positive relationship with adult attachment avoidance (r=

.18, p<0.01). Pet attachment avoidance was negligibly positively associated with adult avoidant

attachment (r= .09, p<0.01) and was not significantly associated with adult anxious attachment

(r= .01, p>0.05). The results indicated that as pet owners’ adult attachment styles became more anxious (and therefore less secure), their pet attachment styles also became more anxious; the same was not true for avoidance.

In constructing the original, Hebrew-language version of the PAQ, the instrument utilized in the current study, Zilcha-Mano, Mikulincer, and Shaver (2011) found that pet attachment

anxiety had a strong positive relationship with adult attachment anxiety (r= .60, p<0.001) and a moderate positive relationship with adult avoidant attachment (r= .08, p<0.01). They also

reported that pet attachment avoidance was not significantly associated with adult avoidant

attachment (r= .08, p>0.05) but it was negligibly positively associated with adult anxious

attachment (r= .08, p<0.01).

The findings in the current study, similar to Zilcha-Mano et al., are somewhat consistent

with a matching hypothesis, which would suggest that attachment styles may be similar for pets

and for romantic relationships, given that the correlations exist and are statistically significant,

164

though “far from perfect” (p. 350). Adult attachment anxiety was significantly related to both pet

attachment style orientations, suggesting that individuals with more anxious attachment styles in

their adult relationships tended to express more insecure attachments to their pets. However,

given the negligible correlation between adult attachment avoidance and pet attachment

avoidance, one should not assume that the relationships represent equivalent constructs. In

considering practical implications of these results, one could assume that avoidant attachment in

relationships with pets was not associated adult attachment style in a meaningful way, so the

relationship between adult owners and their avoidant orientation to pets, when present, is

separate and distinct from adult relationship attachment. Perhaps the avoidance component is

less salient for pet-owner relationships, or avoidance could be less trait-like than anxiety in pet

attachment. Findings from the current study suggest that the sample used was similar in response

style and patterns to previous research (Zilcha-Mano et al., 2011).

Hypothesis 9. I hypothesized that higher levels of secure pet attachment (defined as low

anxiety and low avoidance) would be positively correlated with subjective well-being and perceived social support in the ninth hypothesis. Some sub-components of this hypothesis were supported and statistically significant. Pet attachment avoidance was significantly negatively correlated with positive affect (r= -.09, p<0.05) but was not significantly correlated with perceived social support (r= -.03, p>0.05), satisfaction with life (r= .03, p>0.05), and negative affect (r= .04, p>0.05). Pet attachment anxiety was significantly positively correlated with negative affect (r= .32, p<0.01) and was significantly inversely correlated with perceived social support (r= -.16, p<0.01), satisfaction with life (r= -.22, p<0.01), and positive affect (r= -.11

p<0.01). The strength of the relationships varied from negligible to moderate. In reviewing the

practical significance of the results, one would note that adult pet owners’ attachment avoidance

165 in their relationship to a pet is not meaningfully related to his or her sense of social support or subjective well-being. Pet attachment anxiety is related to these concepts. As pet owners become more anxious in their attachments to their pets, they are more likely to report diminished social support and lower overall well-being. It should also be noted that pet attachment avoidance scores were very low (M=19.50, SD= 6.59) in comparison to pet attachment anxiety scores

(M=32.86, SD= 13.33), indicating that overall, participants were not avoidant in their relationships with their pets.

The results of portions of the ninth hypothesis confirm and support previously published research which suggests that pet attachment relationships can provide social support

(Antonacopoulos & Pychyl, 2010; Yorke et al., 2008) and are linked with well-being (Clines,

2010; Virues-Ortega & Buela-Casal, 2006; Zilcha-Mano et al., 2011). Results from the current study indicate that one particular aspect of secure pet attachment orientation, defined by lower levels of anxious attachment, are associated with high levels of perceived social support and an increased sense of subjective well-being in pet-owning adults. These correlations ranged from negligible to moderate, and all were statistically significant. The results suggest that an anxious pet attachment style is meaningfully related to satisfaction with life and with perceived social support, which could have practice implications in counseling psychology. Perceived social support, adult attachment style, and subjective well-being are constructs which are often assessed by counselors. Asking about relationships with companion animals could provide additional information about a client’s experiences and worldview. Understanding possible interrelationships between forms of attachment, one’s sense of social support, and self-reported well-being could also be useful identifying effective interventions for clients or when conceptualizing client concerns.

166

Hypothesis 10. In the tenth hypothesis, I expected pet attachment style to significantly predict subjective well-being after controlling for adult attachment. The hypothesis was supported in two of the six paths tested. Pet attachment avoidance significantly predicted satisfaction with life, and pet attachment anxiety significantly predicted negative affect. When combined in the regression models, overall adult attachment style predicted all three subjective well-being outcomes in a manner that was statistically significant but represented little practical meaning.

As previously stated, two components of the tenth hypothesis were supported, even after controlling for the demographic variables of age, gender, relationship status, and education level and for adult attachment styles. Pet attachment (combining both anxiety and avoidance) accounted for 6.3% of the variance in positive affect, 16.9% of the variance in negative affect,

and 23.4% of the variance in satisfaction with life, and all were statistically significant at the p<0.001 level. In contrast to what was expected, neither aspect of pet attachment was a predictor of positive affect, meaning that an individual’s pet attachment style did not have a significant impact on his or her positive emotionality. Adding pet attachment orientation to the model only explained an additional 0.4% of the variance in positive affect above and beyond demographic variables and adult attachment. Also in contrast to what was expected, only one aspect of pet attachment (anxious pet attachment) was a predictor of negative affect, meaning that an individual’s pet attachment anxiety had a statistically significant impact on his or her negative emotionality. Adding pet attachment orientation to the model explained an additional 2.0% of the variance in negative affect above and beyond demographic variables and adult attachment. Again in contrast to what was expected, only one aspect of pet attachment (avoidant pet attachment) was a predictor of satisfaction with life, meaning that an individual’s pet attachment avoidance

167

had a statistically significant and positive impact on his or her overall sense of life satisfaction.

Adding pet attachment orientation to the model explained an additional 0.8% of the variance in

negative affect above and beyond demographic variables and adult attachment. In deriving

practical meaning from these results, one would note that a pet owner’s attachment to his or her

pet could have an impact on his or her sense of well-being, but other factors likely have a more

apparent effect on emotionality and life satisfaction.

These findings indicated that pet attachment style has some predictive role on subjective

well-being, and it does so in ways which are statistically but perhaps not practically meaningful

(See Tables 14 and 15). There are several possible explanations for the mixed findings in this

hypothesis. First, as mentioned previously, many of the respondents reported low levels of pet

attachment anxiety and avoidance, as shown by the descriptive statistics in Table 1, which could suggest bias or non-heterogeneity in the sample. Additionally, previous researchers, in acknowledging their own biases, have noted that much of the published data tends to over- emphasize the positive aspects of the human-animal bond, without investigating trends in response patterns (Beck, 1985).

The results of the tenth hypothesis, which was exploratory in nature, suggest that it may be helpful to learn more about individuals’ relationships with their companion animals when attempting to increase their overall sense of subjective well-being. Regardless of age, gender, relationship status, educational background, or adult attachment style, it may be possible to positively impact an adult’s sense of subjective well-being by increasing the sense of security, comfort, and solace that the person experiences in his or her relationship to a companion animal.

By enhancing an individual’s sense of attachment to his or her companion animal, or by augmenting the sense that relationships with a companion animal can provide a secure base or

168

safe haven, for example, the person may then experience a significant decrease in negative affect

or an increase in his or her positive affect or satisfaction with life.

Hypothesis 11. In the eleventh hypothesis, which was exploratory, I expected that perceived social support would significantly predict pet attachment style after controlling for adult attachment. This hypothesis was not supported because perceived social support was not found to be a significant predictor in either pet attachment avoidance or pet attachment anxiety, despite the overall models being significant.

After controlling for the demographic variables of age, gender, relationship status, education level, and adult attachment (both anxiety and avoidance) perceived social support accounted for 4.3% of the variance in pet attachment avoidance and 28.8% of the variance in pet attachment anxiety, and both were statistically significant at the p<0.001 level. Though the overall models were statistically significant, it is also important to consider the role of perceived social support in each outcome of pet attachment. In particular, the ΔR² was .000 for both pet attachment anxiety and pet attachment avoidance, meaning that adding perceived social support to the model explained no additional variance in pet attachment orientation (See Tables 16 and

17). In contrast to what was expected, perceived social support was not a predictor of pet attachment, meaning that an individual’s sense of perceived social support did not explain additional variance in pet attachment.

There are several possible explanations for the varied findings in this hypothesis. First, as mentioned previously, many of the respondents endorsed being securely attached to their companion animals (reflecting low levels of pet attachment anxiety and avoidance), as shown by the descriptive statistics in Table 1. Additionally, as shown in Table 2, not all of the correlations between constructs were found to be statistically significant and those that were statistically

169

significant were not strongly correlated, which made it unlikely for the regression analyses to

become significant. It is also possible that the instrument used to assess pet attachment did not

fully do so. At the time the survey was administered, it was the first time the English version of

Zilcha-Mano’s PAQ had been utilized. Though unlikely, it is possible that the instrument has

challenges with internal or external validity, given that this was the inaugural administration and

interpretation of the English version of the PAQ. Beyond these explanations, it is also possible that the attachment relationship between an adult and his or her companion animal is somewhat insulated from the influence of perceived social support. Unlike romantic attachment relationships, which are interpersonal and social in nature, the relationship between an individual and his or her pet may not leave the confines of one’s living space. As a result, it is possible that the attachment bond is unique in a person’s life, and perceived social support from family, friends, and significant others may have little causal role in the dynamics and quality of the attachment bond that an individual experiences toward a companion animal. Additionally, the analyses were not completed for type of pet owned; it is possible that owners responded differentially based on whether their attachment was to a cat, horse, or dog, for example.

Hypothesis 12. An additional goal of this study was to investigate whether the effects of

perceived social support on subjective well-being would be partially mediated by pet attachment

style, above and beyond adult attachment. Hypothesis 12 was not supported, because perceived

social support was not a significant predictor of pet attachment style. This suggests that other

variables may be responsible for an individual’s pet attachment style, and lends support to the

concept that pet attachment is a separate and distinct entity from adult attachment style, an

argument which has been suggested by previous researchers (Antonacopoulos & Pychyl, 2010;

Yorke et al., 2008; Zilcha-Mano et al., 2011).

170

Hypothesis 13. Another goal of the current study was to explore whether lower levels of perceived social support would strengthen or moderate the association between pet attachment and subjective well-being after controlling for adult attachment. This was the purpose of the 13th hypothesis, which was exploratory in nature.

This hypothesis was not supported, because no significant moderation effects or changes in R2 were found for the interaction between adult attachment and perceived social support in predicting participants’ subjective well-being scores. Consistent with the regression results in

Tables 21and 22, and as shown by the path coefficients in Figure 14, lower levels of perceived social support did not have a significant effect on the causal relationship between pet attachment style (PAtt, measured by pet attachment avoidance and pet attachment anxiety) and subjective well-being (SWB, measured by positive affect, negative affect, and satisfaction with life). These findings suggest that perceived social support does not significantly impact the relationship between pet-owning adults’ pet attachment styles and their subjective well-being.

Applied Implications

Attachment theory is widely recognized and applied to the exploration of human interpersonal relationships, and in recent years, some researchers have established empirical support for connections between attachment style and perceived social support (Cohen & Wills,

1985; Deci et al., 2006; MacDonald & Leary, 2005; Sadava et al., 2009; Thoits, 1986). The results of the current study support and extend these findings. In a sample of English-speaking, pet-owning adults, perceived social support was significantly correlated with adult attachment style, and it was a significant predictor of adult attachment orientation. As suggested in the study’s results, regardless of age, gender, relationship status, or educational background, it may be possible to increase the security and health of an individual’s attachment style by increasing

171

that person’s sense of available social support from friends, family members, and significant

others.

Additionally, researchers have also established empirical support linking attachment style

to health outcomes and self-reported well-being (Sadava et al., 2009; Siegel, 2001; Strathearn,

2011). The results of the current study suggest that, in a sample of English-speaking, pet-owning

adults, adult attachment style was significantly correlated with self-reported measures of subjective well-being, and adult attachment orientation was a significant predictor of subjective well-being. In applied terms, the experience of secure attachment led to enhanced well-being.

Findings from the present study also suggest that adult attachment style is a potential mediator of the relationship between perceived social support and subjective well-being. From a clinical perspective, knowing how attachment orientation and one’s support system could interact with and affect well-being may be particularly useful to psychologists and other mental health professionals who are working with clients in need. Incorporating psychosocial support or important interpersonal relationships into a treatment plan could have a significant effect on symptom reduction or enhance a client’s sense of overall well-being. From an applied perspective, the finding that perceived social support from friends predicted subjective well- being even after controlling for adult attachment has important implications. It provides support for considering the inclusion of one’s social support system in order to increase their overall satisfaction with their lives.

On a deeper level, this research suggests that secure attachment relationships can mediate the relationship between social support and subjective well-being, such that the presence of secure attachment can reduce the negative impact of perceived social support on diminished well-being. If, for example, a client indicated that he or she felt little support from family,

172

friends, or significant others, a therapist could buffer this potentially negative impact on overall

well-being by providing a safe, trusting, therapeutic alliance. In other words, the presence of a

secure attachment (from a healthy relationship with an individual therapist or from meaningful

participation in a therapy group, for example) can have a beneficial impact when a person is

experiencing low levels of social support.

A primary aim of the current study was to establish and provide empirical support for the

construct of pet attachment. The results in this area were mixed, and worthy of additional

exploration in the future. Pet attachment anxiety and pet attachment avoidance emerged as

meaningful constructs which were distinct from adult attachment anxiety and adult attachment

avoidance. The results suggest that an anxious pet attachment style is meaningfully related to

satisfaction with life and with perceived social support, which could have practice implications

in counseling psychology. Perceived social support, adult attachment style, and subjective well-

being are constructs which are often assessed by counselors. In clinical settings, asking about

relationships with companion animals could provide additional information about a client’s

experiences and worldview. Understanding possible interrelationships between forms of

attachment to living beings, one’s sense of social support, and self-reported well-being could also be useful identifying effective interventions for clients or when conceptualizing their concerns. In investigating causal relationships between pet attachment, subjective well-being, and perceived social support, the study found that, beyond demographic characteristics and adult attachment styles, pet attachment was a meaningful construct, which could have practical implications. One clinical implication of this research suggests that it may be possible to positively impact an adult’s sense of subjective well-being by increasing the sense of security, comfort, and solace that the person experiences in his or her relationship to a companion animal.

173

By enhancing an adult’s sense of attachment to his or her companion animal, or by augmenting

the sense that relationships with a companion animal can provide a secure base or safe haven, for

example, the individual may then experience a significant decrease in negative affect or an increase in his or her positive affect or satisfaction with life. In contrast to what was predicted, perceived social support did not have a causal role in determining pet attachment, and pet attachment did not mediate or moderate the relationship between social support and subjective well-being.

Limitations and Future Directions

The present results should be interpreted in light of the study’s limitations. Although efforts were made to recruit a diverse sample of pet-owning adults in terms of gender identity, cultural background, geographic location, relationship status, education level, sexual orientation, and spiritual affiliation, the study sample was somewhat homogeneous and self-selected. The majority of the participants were women living in the United States who identified themselves as being at least partially Caucasian. The sample was also heavily straight in sexual orientation, highly educated, and tended to be young. Although attempts were made to control for the potential influence of age, gender, relationship status, and educational background, the generalizability of the results may be impacted by the characteristics of the participants. Further, the sample was both young and educated, and thus, could be perceived to be highly functioning.

Additionally and as previously noted, the sample was rather secure in both their adult

attachment and their pet attachment scores. Additionally, participants endorsed feeling highly

supported by friends, family and significant others, and they expressed high levels of subjective

well-being. All participants volunteered their participation in the study because there was no

random sampling. Numerous participants contacted me after completing the surveys, asking for

174

results and expressing gratitude for their companion animals. Though unlikely, based on the

mixed results on the findings about pet attachment outcomes, it is possibly that the participants

overemphasized positive aspects of their bonds with their selected companion animal, which

could alter the results. Several participants also contacted me and informed me that, despite

being asked to select one companion animal when completing the PAQ, they chose to answer the

questions by considering all of their pets. This may have also had a detrimental impact on the

results because pet attachment, like attachment theory, focuses on the nuanced connection

between a person and one other living being.

Although attachment theory is a universal concept, perhaps including individuals from a greater range of geographic locations or who are more diverse in terms of age, gender identity, sexual orientation, relationship status, or educational level would provide more external validity to the results. Increasing the heterogeneity of the sample could facilitate a richer analyses of the various impacts that perceived social support could have on SWB across demographic variables.

In particular, to assist with generalizing these findings to English-reading pet owners across the lifespan, future studies should include a greater range of ages and older participants.

Future research in this area of research should continue exploring the role that attachment to a companion animal may have on an individual’s functioning. Again, this research was exploratory in nature, and the instrument used to assess pet attachment (Zilcha-Mano’s Pet

Attachment Questionnaire, PAQ) was used for the first time in its English version. Replication of the study, with this instrument or with other instruments which assess aspects of the unique connection between a person and his or her companion animal, is needed prior to generalizing results or making broad statements about the impact of pet attachment on one’s functioning.

Additional analyses could assess differences in forms of perceived social support beyond a

175

global total. It is possible that the support received from friends, family members, or significant

others is different, or that stressors could exist within the relationships between individuals and

their families, friends, or romantic partners. Such stressors could be assessed by examining the

perceived social support from each of these types of relationships, rather than assessing the

global total, as was completed in the current study. Additionally, further analyses could also

examine the impact, if any, that the type of pet has on results. It is possible that a difference

exists for attachment relationships between dogs and their owners and cats and their owners, for

example.

In sum, including a more heterogeneous sample from a more diverse group would

broaden the generalizability of the findings. Exploring other explanations for mediators or

moderators of the relationship between perceived social support and subjective well-being seems necessary. This study highlighted the significant role of adult attachment in this link, but it is necessary to gather more information about the unique potential role of pet attachment. It might also be of specific interest to examine the role of the mode of perceived social support as a mediator or moderator variable rather than the predictor variable. This group of English-reading companion animal owners may represent a narrow range of life satisfaction in that the majority of participants were highly educated and as a result they may exhibit a skewed representation of subjective well-being (i.e., higher levels of life satisfaction). Therefore, future studies should aim to use a global assessment of general life satisfaction and in core areas of life (e.g., home, work, interpersonal relationships). Lastly, examining the unique contribution of social support from friends, significant others, and family would assist with better understanding whether the relationship between attachment and well-being was influenced by a specific type of support, or lack thereof.

176

Further, it is important to acknowledge that the conceptualization and psychometric

properties of pet attachment may vary. In particular, pet attachment measures have only recently been constructed with a basis in theory, and the English version of the instrument had not yet been validated prior to use. It is possible that problems with the measurement of pet attachment may have influenced the results of this study.

Conclusion

As previously noted, researchers have found physiological evidence (e.g., Baun et al.,

1984; Cole et al., 2007; Lust et al., 2007; Miller et al., 2009) and psychosocial support

(Antonacopoulos & Pychyl, 2010; Knight & Edwards, 2008; Yorke, Adams, & Coady, 2008) for

benefits of having an emotional connection with a companion animal. Many early attempts to

observe and understand the human-animal bond (HAB) were anecdotal and did not rely on a

theory through which to understand the connection between person and pet (Jessee, 1982,

Levinson, 1965). Other research on these effects has been atheoretical in nature. Scholars

completed experiments, gathered data, and performed statistical analyses without using a theory

to drive their investigation (Esteves & Stokes, 2008; Flynn, 2000). Additionally, there has been a

lack of focus on the role of moderating or mediating variables such as pet attachment in better

understanding the relationship between adult attachment, daily functioning, and mental health.

In more recent years, scholars and researchers have begun to use terms from attachment

theory in efforts to capture aspects of the HAB (e.g., Cohen, 2002; Marks, Keopke, & Bradley,

1994; Sable, 2004; Yorke, 2010). Zilcha-Mano, Mikulincer, and Shaver (2011) have successfully

applied attachment theory to the HAB in Israel, and their self-report measure of pet attachment is

now available in English. The Pet Attachment Questionnaire (PAQ, Zilcha-Mano, Mikulincer, &

Shaver, 2011) assesses the HAB in terms of avoidant and anxious pet attachments and provides a

177 reliable measure through which to understand a human’s connection to his or her pet. The current study addressed gaps in the literature through the use of an organized theoretical framework of pet attachment as it pertains to perceived social support and subjective well-being. A primary goal was to gain additional insight about the contribution of pet attachment to people's well- being and perceived social support, above and beyond interpersonal attachment. The results of the current study suggest that pet attachment is related to but also distinct from perceived social support, subjective well-being, and adult attachment. Further, the results suggest that pet attachment may have a meaningful impact on pet owners’ subjective well-being.

178

References

Adamle, K.N., Riley, T.A., Carlson, T. (2009). Evaluating college student interest in pet therapy. Journal of American College Health, 57, 545-548. Aiken, L. S. & West, S. G. (1991). Testing and interpreting interactions in multiple regression. Thousand Oaks, CA: Sage. Ainsworth, M., Blehar, M., Waters, E., & Wall, S. (1978). Patterns of Attachment. Hillsdale, NJ: Erlbaum. Albert, A. & Bulcroft, K. (1988). Pets, families, and the life course. Journal of and Family, 50, 543-552. Antonacopoulos, N. M. D. & Pychyl, T. A. (2010). The possible role of companion-animal anthropomorphism and social support in the physical and psychological health of dog guardians. Society and Animals, 18,379-395. Astrup, C. W., Gantt, W. H., & Stephens, J. H. (1979). Differential effects of person in the dog and in the human. The Pavlovian Journal of Biological Science, 14, 104-107. Allen, K., Blascovich, J., Tomaka, J., & Kelsey, R. (1991). Presence of human friend and pet dogs as moderators of autonomic responses to stress in women. Journal of Personality and Social Psychology, 61(4), 582-589. Bagley, D. K. & Gonsman, V. L. (2005). Pet attachment and personality type. Anthrozoos, 18, 28-42. Barker, R. T. (2005). On the edge or not? Opportunities for interdisciplinary scholars in business communication to focus on the individual and organizational benefits of companion animals in the workplace. Journal of Business Communication, 42, 299315. Doi: 10.1177/0021943605277399. Baun, M., Bergstrom, N., Langston, N., & Thoma, L. (1984). Physiological effects of human/ companion animal bonding. Nursing Research, 33(3), 126-129. Beatson, J. & Taryan, S. (2003). Predisposition to depression: The role of attachment. Australian and New Zealand Journal of Psychiatry, 37, 219–225. Beck, A.M. (1985). The therapeutic use of animals. The Veterinary Clinics of North America. Small Animal Practice, 15, 365-75. Beck, A.M. & Katcher, A.H. (1984). A new look at pet-facilitated therapy. Journal of the American Veterinary Medical Association, 184, 414-21. Beck, A.M., Seraydarian, L., & Hunter, F. (1986). The use of animals in the rehabilitation of psychiatric inpatients. Psychological Reports, 58, 63-6. Beck, L. & Madresh, E. A. (2008). Romantic partners and four-legged friends: An extension of attachment theory to relationships with pets. Antrhozoos, 21, 43-56.

179

Beetz, A., Kotraschal, K., Turner, D. C., Hediger, K., Uvnas-Moberg, K., & Julius, H. (2011). The effect of a real dog, toy dog, and friendly person on insecurely attached children during a stressful task: An exploratory study. Antrhozoos, 24, 349-368. DOI: 10.2752/175303711X13159027359746. Beetz, A. Uvnäs-Moberg, K, Julius, H. & Kotrschal, K. (2012). Psychosocial and psychophysiological effects of human-animal interactions: The possible role of oxytocin. Frontiers in Psychology, 3, 1-15. Doi: 10.3389/fpsyg.2012.00234. Berryman, J. C., Howells, K., & Lloyd-Evans, M. (1985). Pet owner attitudes to pets and people: A psychological study. Veterinary Record, 117, 659-661. Bowlby, J. (1970). Disruption of affectional bonds and its effects on behavior. Journal of Contemporary Psychotherapy, 2, 75-86. Bowlby, J. (1979). The making and breaking affectional bonds. London: Tavistock. Bowlby, J. (1980). By ethology out of psycho-analysis: An experiment in interbreeding. Animal Behavior, 28, 649-656. Bowlby, J. (1982). Attachment and loss: Vol. 1. Attachment. New York: Basic Books. Brasic, J.R. (1998), Pets and health. Psychological Reports, 83, 1011-1024. Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measurement of adult romantic attachment: An overview. In J. A. Simpson & W.S. Rholes(Eds.), Attachment theory and close relationships (pp. 46-76). New York: Guilford Press. Brickel, C.M. (1982). Pet-facilitated psychotherapy: A theoretical explanation via attention shifts. Psychological Reports, 50, 71-74. Brodie, S. J. & Biley, F. C. (1999). An exploration of the potential benefits of pet-facilitated therapy. Journal of Clinical Nursing, 8, 329-337. Brown, S. (2003). Ethnic variation in pet attachment among students at an American school of veterinary medicine. Society & Animals, 11, 101-102. Brown, S. & Katcher, A. (2001). Pet attachment and dissociation. Society & Animals, 9, 25-41. Cangelosi, P.R. & Embrey, C.M. (2006). The healing power of dogs: Cocoa’s story. Journal of Psychosocial Nursing, 44, 17-20. Cassidy, J. (1999). The nature of the child’s ties. In J. Cassidy & P. R. Shaver (Eds.), Handbook of attachment: Theory, research, and clinical applications (4-20). New York: Guilford. Cavanaugh, L. A., Leonard, H. A., & Scammon, D. L. (2008). A tail of two personalities: How canine companions shape relationships and well-being. Journal of Business Research, 61, 469-479. Chinner, T. L. & Dalziel, F. R. (1991). An exploratory study on the viability and efficacy of a pet-facilitated therapy project within a hospice. Journal of Palliative Care, 7, 13-20.

180

Cline, K. M. C. (2010). Psychological effects of dog ownership: Role strain, role enhancement, and depression. The Journal of Social Psychology, 150, 117-131. Coan, J. A. (2010). Adult attachment and the brain. Journal of Social and Personal Relationships, 27, 210-217. DOI: 10.1177/0265407509360900 Coan, J. A., Schaefer, H. S., & Davidson, R. J. (2006). Lending a hand: Social regulation of the neural response to threat. Psychological Science, 17, 1032-1039. Cohen, S. & McKay, G. (1984). Social support, stress, and the buffering hypothesis: A theoretical analysis. In A. Baum, S. E. Taylor, & J. E. Singer (Eds.), Handbook of psychology and health (pp. 253-267). Hillsdale, NJ: Erlbaum. Cohen, S. & McKay, G. (1984). Social support, stress, and the buffering hypothesis: A theoretical analysis. In A. Baum, S. E. Taylor, & J. E. Singer (Eds.), Handbook of psychology and health (pp. 253-267). Hillsdale, NJ: Erlbaum. Cohen, S. P. (2002). Can pets function as family members? Western Journal of Nursing Research, 24, 621-638. DOI: 10.1177/019394502320555386 Cole, K. M., Gawlinski, A., Steers, N., & Kolterman, J. (2007). Animal-assisted therapy in patients hospitalized with heart failure. The American Journal of Critical Care, 16, 575- 585. Connors, M. E. (2011). Attachment theory: A “secure base” for psychotherapy integration. Journal of Psychotherapy Integration, 21,348-362. Doi: 10.1037/a0025460 Corson, S.A., Corson, E.O., Gwynne, P.H., & Arnold, L.E. (1975). Pet-facilitated psychotherapy in a hospital setting. Current Psychiatric Therapies, 15, 277-86. Cox,R. P. (1993). The human/animal bond as a correlate of family functioning. Clinical Nursing Research, 2, 224-231. Crawford, E. K., Worsham, N. L., & Swinehart, E. R. (2006). Benefits derived from companion animals, and the use of the term “attachment.” Anthrozoos, 19, 98-112. Davis, J.H. (1988). Animal-facilitated therapy in stress mediation. Holisitc Nursing Practice, 3, 75-83. Davis, J.H. & Juhasz, A.M. (1984). The human/companion animal bond: How nurses can use this therapeutic resource. Nursing and Healthcare,5, 497-501. Deci, E. L., La Guardia, J. G., Moller, A. C., Scheiner, M. J., & Ryan, R. M. (2006). On the benefits of giving as well as receiving autonomy support: Mutuality in close friendships. Personality & Social Psychology Bulletin, 32, 313-327. Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with life scale. Journal of Personality Assessment, 49, 71-75. Draper, R.J., Gerber, G.J., & Layng, E.M. (1990). Defining the role of pet animals in psychotherapy. Psychiatric Journal of the University of Ottawa, 15, 169-172.

181

Edens, J. ., Larking, K. T., & Abel, J. L. (1992). The effect of social support and physical touch on cardiovascular reactions to mental stress. Journal of Psychosomatic Research, 36, 371-381. Esteves, S. W. & Stokes, T. (2008). Social effects of a dog’s presence on children with disabilities. Anthrozoos, 21, 5-15. Field, A. (2009). Discovering statistics using SPSS (3rd ed.) Los Angeles: Sage. Finch, J. F., Okun, M. A., Pool, G. J. & Ruehlman, L. S. (1999). A comparison of the influence of conflictual and supportive social interactions on psychological distress. Journal of Personality, 67, 581-622. Flynn, C.F. (2000). Battered women and their animal companions: Symbolic interaction between human and nonhuman animals. Society & Animals, 8, 99-127. Fonagy, P., Luyten, P. & Strathearn, L. (2011). Borderline personality disorder, mentalization, and the neurobiology of attachment. Infant Mental Health Journal, 32, 47-69. DOI: 10.1002/imhj.20283 Frazier, P. A., Tix, A. P., & Barron, K. E. (2004). Testing moderator and mediator effects in counseling psychology research. Journal of Counseling Psychology, 51, 115-134. DOI: 10.1037/0022-0167.51.1.115 Friedmann, E. (1990). The value of pets for health and recovery. Proceedings of the British Small Animal Veterinary Association, 20, 9-17. Friedmann, E., Katcher, A., Lynch, J., & Thomas, S. (1980). Animal companions and one-year survival of patients after discharge from a coronary care unit. Public Health Reports, 95(4), 307-312. Friedmann, E., & Thomas, S. (1995). Pet ownership, social support, and one-year survival after acute myocardial infarction in the cardiac arrhythmia suppression trial (CAST). American Journal of Cardiology, 76, 1213-1217. Friedmann, E., Katcher, A.H., Thomas, S.A., Lynch, J.J., & Messent, P.R. (1983). Social interaction and blood pressure: Influence of animal companions. Journal of Nervous and Mental Disease, 171, 461-5. Friesen, L. (2010). Exploring animal-assisted programs with children in school and therapeutic contexts. Early Childhood Education Journal, 37, 261-267. Doi: 10.1007/s10643-009- 0349-5 Froeschle, J. (2009). Empowering abused women through equine assisted career therapy. Journal of Creativity in Mental Health, 4, 180-190. doi: 10.1080/15401380902945228. Geist, T. S. (2011). Conceptual framework for animal assisted therapy. Child Adolescent Social Work Journal, 28, 243256. Doi: 10.1007/s10560-011-0231-3.

182

Gump, B. B., Polk, D. E., Kamarck, T. W., &Shiffman, S. M. (2001). Partner reactions are associated with reduced blood pressure in the natural environment: Ambulatory monitoring evidence from a healthy, multiethnic adult sample. Psychosomatic Medicine, 63, 423-433. House J. S., Landis K. R., & Umberson, D. (1988). Social relationships and health. Science, 241, 540-544. Hughes, A.E., Crowell, S.E., Uyeji, L., & Coan, J.A. (2012). Developmental neuroscience of borderline pathology: Emotion dysregulation and social baseline theory. Abnormal Child Psychology, 40, 21–33. DOI 10.1007/s10802-011-9555-x Imber-Black, E. (2009). Snuggles, my cotherapist, and other animal tales in life and therapy. Family Process, 48, 459-461. Insel, T. L. & Young, L. J. (2001). The neurobiology of attachment. Nature Reviews, 2, 129-136. Jasperson, J.A. (2010). Animal-assisted therapy with female inmates with mental illness: A case example from a pilot program. Journal of Offender Rehabilitation, 49, 417-433. doi: 10.1080/ 10509674. 2010.499056 Jenkins, J. L. (1986). Physiological effects of petting a companion animal. Psychological Reports, 58, 21-22. Jessee, E.M. (1982). Pet therapy for the elderly. Aging, Sept-Oct, 26-28. Johnson, R.A., Odenaal, J.S.J., & Meadows, R.L. (2002). Animal-assisted interventions research: Issues and answers. Western Journal of Nursing Research, 24, 422-440. Doi: 10.1177/01945902024004009 Johnson, S. B. & Rule, W. R. (1991). Personality characteristics and self-esteem in pet owners and non-owners. International Journal of Psychology, 26, 241-252. Johnson, T P., Garrity, T F and Stalbnes, L. 1992. Psychometric evaluation of the Lexington Attachment to Pets Scale (LAPS). Anthrozoos, 5, 160-176. Kamarck, T. W., Manuck, S. B., & Jennings, J. R. (1990). Social support reduces cardiovascular reactivity to psychological challenge: A laboratory model. Psychosomatic Medicine, 52, 42-58. Kenny, D. (2013, May 4). Moderation. Retrieved from http://davidakenny.net/cm/moderation.htm Kenny, D. (2013, December 16). Mediation. Retrieved from http://davidakenny.net/cm/mediate.htm Knight, S. & Edwards, V. (2008). In the company of wolves: The physical, social, and psychological benefits of dog ownership. Journal of Aging and Health, 20, 437-455. Kobak, R. (2009). Defining and measuring of attachment bonds: Comment on Kurdek (2009). Journal of Family Psychology, 23, 447-449.

183

Kovacs, Z., Kis, R., Rozsa, S., & Rozsa, L. (2004). Animal-assisted therapy for middle-aged schizophrenic patients living in a social institution: A pilot study. Clinical Rehabilitation, 18, 483-486. Krause-Parello, C. A. (2012). Pet ownership and older women: The relationships among loneliness, pet attachment support, human social support, and depressed mood. Geriatric Nursing, 33, 194-203. Kurdek, L. A. (2008). Pet dogs as attachment figures. Journal of Social and Personal Relationships, 25, 247–266. Kurdek, L. A. (2009). Pet dogs as attachment figures for adult owners. Journal of Family Psychology, 23, 439-446. Kwong, M. J. & Bartholomew, K. (2011). “Not just a dog”: An attachment perspective on relationships with assistance dogs. Attachment and Human Development, 13, 421-436. Doi: 10.1080/14616734.2011.584410 Levinson, B.M. (1964). Pets: A special technique in . Mental Hygiene, 48, 243-248. Levinson, B.M. (1965). Pet psychotherapy: The use of pets in the treatment of behavior disorder in childhood. Psychological Reports, 17, 695-698. Lipton, B. & Fosha, D. (2011). Attachment as a transformative process in AEDP: Operationalizing the intersection of attachment theory and affective neuroscience. Journal of Psychotherapy Integration, 21, 253-279. DOI: 10.1037/a0025421 Lust, E., Ryan-Haddad, A., Coover, K, & Snell, J. (2007). Measuring clinical outcomes of animal-assisted therapy: Impact on resident medication usage. The Consultant Pharmacist, 22, 580-585. MacDonald, G. & Leary, M. R. (2005). Why does social exclusion hurt?: The relationship between social and physical pain. Psychological Bulletin, 131, 202-223. Main, M. & Solomon, J. (1990). Procedures for identifying infants as disorganized/disoriented during the Ainsworth strange situation. In Greenberg, M.T., Cicchetti, D., Cummings, E. M. Attachment in the Preschool Years: Theory, Research, and Intervention. Chicago: University of Chicago Press. pp. 121–60. Margolies, L. (1999). The long good-bye: Women, companion animals, and maternal loss. Clinical Social Work Journal, 27, 289-304. Marks, S. G., Koepke, J. E., & Bradley, C. L. (1994). Pet attachment and generativity among young adults. The Journal of Psychology, 128, 641-650. Mason, M.S. & Hagan, C.B. (1999). Pet-assisted psychotherapy. Psychological Reports, 84, 1235-1245.

184

McCandless, P., McCready,K. F., & Knight,L. (1985). A model animal therapy program for mental health settings. Therapeutic Recreation Journal, 19, 55-63. McNicholas, J., Gilbey, A., Rennie, A., Ahmedzai, S., Dono, J., & Ormerod, E. (2005). Pet ownership and human health: A brief review of evidence and issues. British Medical Journal, 331, 1252-1254. Melson, G. F., Kahn, Jr., P. H., Back, A., & Friedman, B. (2009). Robotic pets in human lives: Implications for the human-animal bond and for human relationships with personified technologies. Journal of Social Issues, 65, 545-567. Messent, P.R. (1985). Pets as social facilitators. Veterinary Clinics of North America-Small Animal Practice, 15, 387-393. Michaels, E. (1982). Pets and the elderly: A therapeutic . Canadian Medical Association Journal, 127, 70-71. Mikulincer, M. & Shaver, P. (2007). Attachment in adulthood: Structure, dynamics, and change. New York: Guilford Press. Miller, S. C., Kennedy, C., DeVoe, D., Hickey, M., Nelson, T., & Kogan, L. (2009). An examination of changes in oxytocin levels in men and women before and after interaction with a bonded dog. Anthrozoos, 22, 31-42. Morrison, M.L. (2007). Health benefits of animal-assisted interventions. Complementary Health Practice Review, 12, 51-62. Doi: 10.1177/15332101070302397 Nightingale, F. (1859). Notes on nursing, what it is and what it is not. London: Harrison & Sons. Retrieved from www.countryjoe.com/nightingale.cats.htm on November 12, 2011 Noonan, E. (2008). People and pets. Psychodynamic Practice, 14, 395-407. Odendaal, J., & Lehmann, S. (2000). The role of phenylethylamine during positive human-dog interaction. ACTA Veterinaria, 69(3), 183-188. Ormerod, E. J., Edney, A. T. B., Foster, S. J., & Whyham, M. C. (2005). Therapeutic applications of the human-companion animal bond. The Veterinary Record, 157, 689- 691. Ovesen, L. (2005). Social interaction with dogs. European Journal of Cancer Prevention, 14, 543-544. Pachana, N. A., Ford, J. H., Andrew, B., & Dobson, A., J. (2005). Relations between companion animals and self-reported health in older women: Cause, effect, or artifact? International Journal of Behavioral Medicine, 12, 103-110. Palmer, R. & Custance, D. (2008). A counterbalanced version of Ainsworth’s strange situation procedure reveals secure-base effects in dog-human relationships. Applied Animal Behavior Science, 109, I306319.

185

Parish-Plass, N. (2008). Animal-assisted therapy with children suffering from insecure attachment due to abuse and neglect: A method to lower the risk of intergenerational transmission of abuse? Clinical Child Psychology and Psychiatry, 13, 7-30. Doi: 10.1177/1359104507086338. Parthasarathy, V. & Crowell-Davis, S. L. (2006). Relationship between attachment to owners and separation anxiety in pet dogs (Canis lupus familiaris). Journal of Veterinary Behavior, 1, 109-120. Pavot, W., Diener, E., Colvin, C. R., & Sandvik, E. (1991). Further validation of the satisfaction with life scale: Evidence for the cross-method convergence of well-being measures. Journal of Personality Assessment, 57, 149-161. Pendry, P. & Roeter, S. (2012). Experimental trial demonstrates positive effects of equine- facilitated learning on child social competence. Human-Animal Interaction, 1, 1-19. Pierce, T., Baldwin, M. W., & Lydon, J. E. (1997). A relational schema approach to social support. In G. R. Pierce, B. Lakey, I. G. Sarason, & B. R. Sarason (Eds.), Sourcebook of social support and personality (pp. 19 – 47). New York: Plenum. Prato-Previde, E., Fallani, G., & Valsecchi, P. (2006). Gender differences in owners interacting with pet dogs: An observational study. Ethology, 112, 64-73. Procidano, M. E. & Heller, K. (1983). Measures of perceived social support from friends and family: Three validation studies. American Journal of Community Psychology, 11, 1-24. Ratnasingam, P., & Bishop, G. D. (2007). Social support schemas, trait anger, and cardiovascular responses. International Journal of Psychophysiology, 63, 308 – 316. Riedl, R. & Javor, A. (2012). The biology of trust: Integrating evidence from genetics, endocrinology, and functional brain imaging. Journal of Neuroscience, Psychology, and Economics, 5, 63-91. DOI: 10.1037/a0026318 Risley-Curtiss, C. (2010). Social work practitioners and the human-companion animal bond: A national study. Social Work, 55, 38-46. Rogers, C. (1961). On becoming a person. Houghton Mifflin: Boston. Rogers, C. (1980). A way of being. Houghton Mifflin: Boston. Roisman, G. I., Holland, A., Fortuna, K., Fraley, R. C., Clausell, E., & Clarke, A. (2007) The adult attachment interview and self-reports of attachment style: An empirical rapprochement. Journal of Personality and Social Psychology, 92, 678–697. Doi: 10.1037/0022-3514.92.4.678 Ryder, E. L. (1985) Pets and the elderly: A social work perspective. The Veterinary Clinics of North America, Small Animal Practice, 15, 333-343. Rynearson, E. K. (1978).Humans and pets and attachment. British Journal of Psychiatry, 133, 550-555.

186

Sable, P. (1995). Pets, attachment, and well-being across the life cycle. Social Work, 40, 334- 341. Sable, P. (2004). Attachment, ethology, and adult psychotherapy. Attachment & Human Development, 6, 3-19. Doi: 10.1080/14616730410001663498 Sacks, A. (2008). The therapeutic use of pets in private practice. British Journal of Psychotherapy, 24, 501-521. Sadava, S.W., Busseri, M.A., Molnar, D.S., Perrier, C.P.K., & DeCourville, N. (2009). Investigating a four-pathway model of adult attachment orientation and health. Journal of Social and Personal Relationships, 26, 604-633. Doi: 10.1177/0265407509354402\ Schore, A. N. (2001). Effects of a secure attachment relationship on right brain development, affect regulation, and infant mental health. Infant Mental Health Journal, 22, 7-66. Scott, J. P. & Fuller, J. L. (1965). The development of social relationships. In: Genetics and the social behavior of the dog. University of Chicago Press: Chicago. Sellers, D. M. (2006). The evaluation of an animal assisted therapy intervention for elders with dementia in long-term care. Activities, Adaptation, & Aging, 30, 61-77. Shiloh, S., Sorek, G., & Terkel, J. (2003). Reduction of state-anxiety by petting animals in a controlled laboratory experiment. Anxiety, Stressk and Coping, 16, 387-395. Shore, E. R., Douglas, D. K., & Riley, M. L. (2005). What’s in it for the companion animal? Pet attachment and college students’ behaviors toward pets. Journal of Applied Animal Welfare Science, 8, 1-11 Siegel, D.J. (2001). Toward an interpersonal neurobiology of the developing mind: Attachment relationships, “mindsight,” and neural integration. Infant Mental Health Journal, 22, 67- 94. Smith, T. W., Ruiz, J. M., & Uchino, B. N. (2004). Mental activation of supportive ties, hostility, and cardiovascular reactivity to laboratory stress in young men and women. Health Psychology, 23, 476 – 485. Sockalingham, S., Li, M., Krishnadev, U., Hanson, K., Balaban, K., Paciones, L. R., & Bhalero, S. (2008). Use of animal-assisted therapy in the rehabilitation of an assault victim with a concurrent mood disorder. Issues in Mental Health Nursing, 29, 73-84. Doi: 10.1080/01612840701748847 Speck, R.V. (1964). Mental health problems involving the family, the pet, and the veterinarian. American Veterinary Medical Association Journal, 145, 150-4. Sperling, M. B., Foelsch, P, & Grace, C. (1996). Measuring adult attachment: Are self-report instruments congruent? Journal of Personality Assessment, 67, 37-51. Stone, G. L. (1980). A cognitive-behavioral approach to counseling psychology: Implications for practice, research, and training. Praeger: New York.

187

Strathearn, L. (2011). Maternal neglect: Oxytocin, dopamine, and the neurobiology of attachment. Journal of Neuroendocrinology, 23, 1054-1065. doi: 10.1111/j.1365- 2826.2011.02228.x Taylor, N. & Signal, T. D. (2005). Empathy and attitudes to animals. Anthrozoos, 18, 18-27. Thoits, P. A. (1986). Social support as coping assistance. Journal of Clinical and Consulting Psychology, 54, 416-423. Tilsen, J. (1998). Vignettes from the doghouse: Tales of canine co-therapy. Journal of Family Life, 4, 53-55. Topal, J., Gacsi, M., Miklosi, A., Viranyi, Z., Kubinyi, E., &Csanyi, V. (2005). Attachment to humans: A comparative study on hand-reared wolves and differently socialized dog puppies. Animal Behaviour, 70, 1367-1375. Topal, J., Miklosi, A., Csanyi, V., & Doka, A. (1998). Attachment behavior in dogs (Canis familiaris): A new application of Ainsworth’s (1969) strange situation test. Journal of Comparative Psychology, 112, 210-229. Tsai, C., Friedmann, E., & Thomas, S. A. (2010). The effect of animal-assisted therapy on stress responses in hospitalized children. Anthrozoos, 23, 245-258. Tuber, D. S., Sanders, S., Hennessy, M. B., & Miller, J. A. (1996). Behavioral and glucocorticoid responses of adult domestic dogs (Canis familiaris) to companionship and social separation. Journal of Comparative Psychology, 110, 103-108. Virues-Ortega, J. & Buela-Casal (2006). Psychophysiological effects of human-animal interaction: Theoretical issues and long-term interaction effects. The Journal of Nervous and Mental Disease, 194, 52-57. Doi: 10.197/.01nmd.0000195354.03653.63 Walsh, F. (2009). Human-animal bonds I: The relational significance of companion animals. Family Process, 48, 462-480. Walsh, F. (2009). Human-animal bonds II: The role of pets in family systems and family therapy. Family Process, 48, 481-499. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063-1070. Wells, D. L. (2009). The effects of animals on human health and well-being. Journal of Social Issues, 65, 523-543. Wiggett-Barnard C. & Steel, H. (2008). The experience of owning a guide dog. Disability and Rehabilitation, 30, 1014-1026. Wilson, C. C. & Barker, S. B. (2003).Challenges in designing human-animal interaction research. American Behavioral Scientist, 47, 16-28. Doi: 10.1177/0002764203255208

188

Wilson, C. C. & Netting, F. E. (1983). Companion animals and the elderly: A state-of-the-art summary. Journal of the American Veterinary Medical Association, 183, 1425-1429. Winefield, H. R., Black, A., & Chur-Hansen, A. (2008). Health effects of ownership of and attachment to companion animals in an older population. International Journal of Behavioral Medicine, 15, 303–310. Doi: 10.1080/10705500802365532. Winefield, H. R., Winefield, A. H., & Tiggemann, M. (1992). Social support and psychological well-being in young adults: The multi-dimensional support scale. Journal of Personality Assessment, 58, 198–210. Woods, S. M. (1965). Psychotherapy and the patient’s pet. Current Psychiatric Therapies, 11, 119-121. Woodward, L. E. & Bauer, A. L. (2007). People and their pets: A relational perspective on interpersonal complementarity and attachment in companion animal owners. Society and Animals, 15, 169-189. Yamauchi, T., Baeyens, M. M., McCoy, J. &Carter, D. (1984). The microflora of animals used in hospital pet therapy: Is there patient risk? Clinical Research, 32, 880. Yorke, J. (2010). The significance of human-animal relationships as modulators of trauma effects in children: A developmental neurobiological perspective. Early Child Development at Care, 180, 559-570. Doi: 10.1080/03004430802181189 Yorke, J., Adams, C., & Coady, N. (2008). Therapeutic value of equine- in recovery from trauma. Anthrozoos, 21, 17-30. Doi: 10.2752/089279308X274038 Zilcha-Mano, S., Mikulincer, M., & Shaver, P.R. (2011). An attachment perspective on human– pet relationships: Conceptualization and assessment of pet attachment orientations. Journal of Research in Personality, 45(4), 345–357. Zilcha-Mano, S., Mikulincer, M., & Shaver, P. R. (2011). Pet in the therapy room: An attachment perspective on animal-assisted therapy. Attachment & Human Development, 13, 541-561. Zilcha-Mano, S., Mikulincer, M., & Shaver, P. R. (in press, 2012). Pets as safe havens and secure bases: The moderating role of pet attachment orientations. Journal of Research in Personality, xx, 10 pages. Zimolag, U.& Krupa, T. (2010). The occupation of pet ownership as an enabler of community integration in serious mental illness: A single exploratory case study. Occupational Therapy in Mental Health, 26, 176-196. Doi: 10.1080/01642121003736101

189

Appendix A

Informed Consent

WASHINGTON STATE UNIVERSITY College of Education The Department of Educational Leadership & Counseling Psychology Research Study Consent Form Study Title: Understanding and Quantifying the Roles of Pet Attachment, Adult Attachment, and Perceived Social Support in Adult Pet Owners' Subjective Well-Being

Researchers: Phyllis Erdman, Ph.D., ELCP Stephanie Wilmore, M. Ed., ELCP Cleveland 160B 4067810848 Pullman, WA 991642114 [email protected] 5093351738 [email protected]

Dear Participant:

You are being asked to take part in a research study carried out by Dr. Phyllis Erdman and Stephanie Wilmore. This form explains the research study and your part in it if you decide to join the study. Please read the form carefully, taking as much time as you need. Contact the researchers to explain anything you don’t understand. You can decide not to participate the study. If you choose to participate, you can change your mind later or quit at any time. There will be no penalty or loss of services or benefits if you decide to not take part in the study or quit later. Based on the Exemption Determination Application submitted for the study titled "Understanding and Quantifying the Roles of Pet Attachment, Adult Attachment, and Perceived Social Support in Adult Pet Owners' Subjective Well-Being.," and assigned IRB # 12894, the WSU Office of Research Assurances has determined that the study satisfies the criteria for Exempt Research at 45 CFR 46.101(b)(2).

What is this study about? The purpose of this study is to investigate the effect of pet attachment on the relationship between levels of perceived social support, adult attachment, and subjective well-being among adults who own pets. You are being asked to take part because you are an adult who currently owns a companion animal. Taking part in the study will take 10 minutes of your time. You cannot take part in this study if you are under 18 years of age, you do not own a pet, or you have significant difficulty understanding the English language.

What will I be asked to do if I am in this study? If you take part in the study, you will be asked to complete the following questionnaires: (a) Experiences in Close Relationships Scale; (b) Pet Attachment Questionnaire; (c) Multidimensional Scale of Perceived Social Support and Pet Social Support; (d) Satisfaction

190

With Life Scale; (e) Positive and Negative Affect Schedule; and (f) a demographic questionnaire. You have the right to refuse to answer any question in the survey, and you may discontinue your participation at any time. If you desire, you may have access to a copy of the data analysis at the end of the study. It will take approximately 10 minutes total in order to complete all the questionnaires.

Are there any benefits to me if I am in this study? The potential benefits to you for taking part in this study are: increased awareness about your relationships with your pet (s) and other adults; increased awareness about your sense of social support and overall well-being;and the opportunity to participate in a study investigating the human-animal bond. Additionally, you may enter your email address into a drawing for prizes, including ten (10) $25 Amazon giftcards. These electronic giftcards are emailed to the email address you provide and your anonymity will be preserved. You do not need to participate in the entire study in order to be entered into the drawing.

Are there any risks to me if I am in this study? The potential risks from taking part in this study are possible increased psychological distress and possible mental fatigue. Will my information be kept private? The data for this study will be kept anonymous to the extent allowed by federal and state law. No published results will identify you, and your name will not be associated with the findings. Data will be coded and a key will be maintained separately. All study data will be kept in a locked filebox within a locked office, and all data entered into a computer will be password protected. The investigators and the Institutional Review Board will have access to the data. The results of this study may be published or presented at professional meetings, but the identities of all research participants will remain anonymous. The data for this study will be kept for five years.

Are there any costs or payments for being in this study? There will be no costs to you for taking part in this study.

Who can I talk to if I have questions? If you have questions about this study or the information in this form, please contact the researcher(s): Phyllis Erdman, Ph.D., ELCP Cleveland 160B Pullman, WA 991642114 5093351738 [email protected]

Stephanie Wilmore, M. Ed., ELCP 4067810848 [email protected]

If you have questions about your rights as a research participant, or would like to report a concern or complaint about this study, please contact the Washington State University

191

Institutional Review Board at (509) 335-3668, or email [email protected], or regular mail at: Albrook 205, PO Box 643005, Pullman, WA 99164-3005.

What are my rights as a research study volunteer? Your participation in this research study is completely voluntary. You may choose not to be a part of this study. There will be no penalty to you if you choose not to take part. You may choose not to answer specific questions or to stop participating at any time.

By selecting the "Next" Button at the bottom of this page, you are giving your consent to participate in this study.

192

Appendix B

Email Recruitment Flyer

Dear Participant:

Pet Attachment, Social Support, and Subjective Well-being

A warm hello to you all! Greetings! My name Stephanie Wilmore and I am a doctoral candidate in the Counseling Psychology program at Washington State University. I am conducting my dissertation research related to pet attachment, social support, and well-being. I would appreciate your help in the completion of this important task. The anonymous survey will take approximately 10 minutes to complete. In return for your time, you can be entered into a drawing for one of ten Amazon gift certificates worth $25.00. I am a member of this division and would like to make sure I get representation from as many diverse individuals and groups as possible.

You are being asked to take part in a research study carried out by Dr. Phyllis Erdman (faculty member and dissertation chair, [email protected]) and Stephanie Wilmore (doctoral candidate, [email protected]). The first page of the link in Survey Monkey explains the research study and your part in it if you decide to join the study. Please read the form carefully, taking as much time as you need. Contact the researchers to explain anything you don’t understand. Based on the Exemption Determination Application submitted for the study titled "Understanding and Quantifying the Roles of Pet Attachment, Adult Attachment, and Perceived Social Support in Adult Pet Owners' Subjective Well-Being," and assigned IRB # 12894, the WSU Office of Research Assurances has determined that the study satisfies the criteria for Exempt Research at 45 CFR 46.101(b)(2).

In order to participate in the study: * You must be an adult (over 18 years of age). * You must currently own at least one pet.

If you are interested in this research, please click the link below. Please also consider forwarding this email/survey to other adult pet owners. https://www.surveymonkey.com/s/BKYXFJR

Your help is greatly appreciated.

193

Appendix C

Demographic Information Questionnaire First, I would like to ask for some general background information about you. Please accurately answer these questions by filling in the blank or selecting the correct response. (All of the following information will be used for research purposes only).

1. Age: ______

2. Country of Residence: ______

3. Country of Birth: ______

4. Gender: 1. Male 2. Female 3. Transgender 4. Other (please specify)______5. Prefer not to answer

5. Which of the following best describes how you identify? 1. Straight/Heterosexual 2. Gay 3. Lesbian 4. Bisexual 5. Asexual 6. Pansexual 7. Other (please specify) ______8. Prefer not to answer

6. Which of the following best describes how you identify (Select all that apply)? 1. White/Caucasian 2. African American/Black 3. Asian-American 4. Native American/American Indian/Alaska Native 5. Latino/a/Hispanic/Spanish-American 6. Native Hawaiian and Pacific Islander 7. Asian-Indian 8. Other (please specify)______

7. Do you have Latina/o Heritage? 1. Yes 2. No

8. What is your relationship status? (Circle number) 1. Single

194

2. Married 3. Cohabitating/Living with partner 4. In a committed relationship, not living together 5. Separated/Divorced 6. Widowed 7. Other (please specify)______

9. Do you have children? Yes/no Number of Children: ______

10. What is the highest level of education that you have completed? (Circle number) 1. Elementary school 2. High School 3. Two-year college 4. Four-year college 5. Graduate School/Professional School 6. Other (please specify)______

11. Please describe your current occupation or job: ______If you are a student, please describe your year in school (Circle number) 1. Undergraduate/College Student 2. Graduate/Professional Student 3. Other (please specify)______

12. What is your religion? (Circle number) 1. Christian (please specify)______2. Jewish 3. Muslim 4. Buddhist 5. Unitarian/Universalist 6. Hindu 7. Native American 8. Sikh 9. Wiccan 10. Pagan 11. Spiritualist 12. Atheist 13. Agnostic 14. Humanist 15. Other (please specify)______16. No Religion 17. Prefer not to answer

195