Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An investigation of anger among adolescents : an attachment perspective Konishi, Chiaki 2009

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata


ubc_2010_spring_konishi_chiaki.pdf [ 2.83MB ]
JSON: 1.0054452.json
JSON-LD: 1.0054452+ld.json
RDF/XML (Pretty): 1.0054452.xml
RDF/JSON: 1.0054452+rdf.json
Turtle: 1.0054452+rdf-turtle.txt
N-Triples: 1.0054452+rdf-ntriples.txt
Original Record: 1.0054452 +original-record.json
Full Text

Full Text

AN iNVESTIGATIONOF ANGER AMONG ADOLESCENTS:AN ATTACHMENTPERSPECTIVEbyCHIAKI KONISHIB.Ed., The Chiba University,1990M.Ed., The Universityof Massachusetts, Amherst,1996M.A., The Universityof British Columbia, 2003A THESIS SUBMITTEDIN PARTIAL FULFILMENTOFTHE REQUIREMENTSFOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATESTUDIES(Human Development,Learning and Culture)THE UNIVERSITY OFBRITISH COLUMBIA(Vancouver)December 2009© Chiaki Konishi, 200911AbstractThe present studyinvestigated the relationshipbetween attachment andanger amongadolescents, examininga hypothesis initiallyproposed by Bowiby (1973)regarding the effects ofadolescents’ attachmentsto parents on angerexperience. ExtendingBowlby’s hypothesis withanother critical angercomponent, angerexpression, a theoretically-refinedmodel was developedand tested. Participantsincluded 776 students (379 boys,397 girls) in grades 8-12. Aspredictedby attachmenttheory, results ofstructural equation modelinganalyses indicated thatadolescents’attachment anxietyand attachment avoidancetoward both mother and fatherfigures werepositively related to theadolescents’ greater levels ofanger intensity. In turn, the increasesin theintensity of angerfeelings were associated withincreases in both anger-in(internalizing) andanger-out (externalizing)expressions. Inaddition, there was a direct effectof attachment anxietyon anger-in expressionbut no direct effects ofattachment anxiety and avoidanceon anger-outexpression. This studyhighlights the importance ofdifferentiating anger dimensionsand thecritical role of anger intensityas a mediator of therelationship between insecure attachmentandanger expressions.Implications of the findingsare further discussed.111TABLE OF CONTENTSAbstractiiTable of ContentsiiiList of TablesviList of FiguresviiiAcknowledgementsixIntroduction1Literature Review4Attachment Theory4Internal workingmodels and traditions ofattachmentresearch 5Measuring attachment:Underlying dimensions ofattachment 12Anger and an AttachmentPerspective17Dimensions and ImpactofAnger to be Considered22Statement of the Problem26Method32Participants32Procedures33Measures33Demographic information33Attachment33Anger34Results37ivData Preparation and Screening37Missing data37Tests ofassumptions38Assessment ofunidimensionality amongstudy variables 44General sex differences on study variables47Tests ofHypotheses50Analyses ofMother FigureAttachment with Entire Sample 51Model test51Tests ofmediation52Analyses ofFather Figure Attachment withEntire Sample 61Model test61Tests ofmediation61Analyses ofMother Figure Attachment by Sex62Model test62Tests ofmediation63Analyses ofFather Figure Attachment by Sex65Model test65Tests ofmediation68Relative impacts ofattachment dimensions andJIgures 69Discussion82Summary and Discussion ofFindings82Implications ofFindings89Limitations92VReferences95Appendices108Appendix A-i: Parental Consent108Appendix A-2: Student Consent110Appendix B: Ethics Approval112Appendix C-i: DemographicInformation Questionnaire 113Appendix C-2: ComprehensiveAdolescent-Parent Attachment Inventory 114Appendix C-3: State-Trait AngerExpression Inventory 119Appendix D: Scree Plots forExploratory Factor Analyses 120Appendix E: Correlation Matricesfor Scale Items 125viList of TablesTable 1. Distribution of Participants by Grade Level and Sex 32Table 2. Sample Sizes for Data Analyses 38Table 3. Results of Normality, and Descriptive Statistics for Study Variables 40Table 4. Intercorrelations Among Study Variables 42Table 5. Results of Factor Analyses for Study Variables 46Table 6. Mean (Standard Deviation) of Predictor and Outcome Scores by Sex 49Table 7. Fit Indices and Standardized Path Coefficients for Models 55Table 8. Direct, Indirect, and Total Effects of Study Variables on Endogenous Variables• 56Table 9. Fit Indices and Standardized Path Coefficients for Models of Father FigureAttachment for Girls 67Table 10. Results of Multicollinearity Diagnosis for Predictor Variables 72Table 11. Intercorrelations Among Study Variables 73Table 12. Summary of Test for Moderation Effects of Sex on the Prediction ofIntensity of Anger (N = 761) 74Table 13. Summary of Test for Moderation Effects of Sex on the Prediction ofAnger-In(N =761) 75Table 14. Summary of Test for Moderation Effects of Sex on the Prediction ofAnger-Out (N =761) 76Table 15. Summary of Regression Analysis for Variables Predicting Intensityof Anger(N =761) 79Table 16. Summary of RegressionAnalysis for Variables PredictingAnger-In (N 761)80Table 17. Summary of RegressionAnalysis for Variables PredictingAnger-Out (N =761)81viiviiiList of FiguresFigure 1. Bartholomew’s (1990) model of adult attachment 10Figure 2. A diagram of anxiety and avoidance in relation to Mainand Solomon’s (1990) infant attachment types (reproducedfrom Brennan, Clark, & Shaver, 1998). 15Figure 3. A model of attachment and anger. 31Figure 4. An attachment- anger model of mother figure with entire sample. . 58Figure 5. An attachment- anger model of father figure with entire sample. 58Figure 6. An attachment- anger model of mother figure with boys only. . 59Figure 7. An attachment- anger model of mother figure with girls only. 59Figure 8. An attachment- anger model of father figure with boys only. . 60Figure 9. An attachment- anger model of father figure with girls oniy. . 60ixAcknowledgementsThrough the process of my dissertationwriting, I have come to re-realize how luckyIhave been, being supported by anumber of people in my life. First, my committeemembers; Dr.Sheila Marshall, I was alwaysstimulated by your challenging and thoughtful feedback. Ialsoenjoyed the unique metaphors you oftenused to describe or explain tenns (for example, a cookiedough to explain the negative aspectof the internal consistency coefficient alphas). They greatlyhelped me for conceptual understanding.Dr. Bruno Zumbo, you have provided me with alearning system in which I become curiousand motivated to apply various methodologicalstrategies to better explain our world.Despite the fact that I was often chasing you around toinquire about statistical questions, you never declinedmy appointments, rather, you weregenerously approachable. Critically, Dr. ShelleyHymel, my research supervisor, certainly youhave become my attachment figurein my graduate-study life. You have taught me to always“think more.” You also never gave upon me. Even when I was feeling down, almost loosing myenthusiastic energy in the research process, you keptencouraging me to move forward.Second, my friends, Jessica Flores, Marc Sasso,Anne Tomlinson, Yuki Tani, Nat RockeHenderson, Rina Bonanno, and Sarah Hickinbottom, you havebeen sincerely supportive in manyways, being my mentors with a cup or many cups ofcoffee and tea, and potato chips.Third, I deeply thank the students who participatedin this research and theadministrators, teachers, staff, and parents, who enthusiastically supportedthis research project.Lastly, I express my immense appreciation to my family members, Ikuko, Hideo, andMasao Konishi, in Japan. Because of you, I was ableto experience such valuable research andlife endeavors at the University of British Columbia,studying abroad. I promise to share withyou, hopefully very soon, the wonderfulexperiences that I have obtained here!IntroductionProblems emanating from adolescents’undercontrolled anger and aggression are amongthe most shared and seriousconcerns of parents, teachers, and educators (McGee, Silva, &Williams, 1983; McWhirter,McWhirter, McWhirter, & McWhirter, 2003; Underwood, 2003).Indeed, clinical and health literaturessuggest that anger is one of the most difficult emotions foradolescents to deal with, and it potentially contributesto many of their physical and mentalhealth problems (Biaggio & Godwin, 1987; Blumberg& Izard, 1985; Chaplin, 2006; Farmer,2002; Moreno, Fuhriman, & Selby, 1993;Pipher, 1994; Riley, Treiber, & Woods, 1989; Robbins& Tanck, 1997; Seidlitz, Fujita,& Duberstain, 2000). Specifically, anger has been associatedwith maladaptive psychologicaloutcomes, including externalizing (Bosworth, Espelage, &Simon, 1999; Conger, Neppl, Kim, & Scaramella,2003; Cornell, Peterson, & Richards, 1999;Helfritz & Stanford, 2006; Swan, Gambone, Fields, Sullivan, & Snow,2005) and internalizingproblems (Blumberg & Izard, 1985; Bridewell& Chang, 1997; Cautin & Overholser, 2001;Chaplin, 2006; Clay, Anderson & Dixon, 1993; Golman& Haaga, 1995; Kopper & Epperson,1996; Newman, Gray, & Fuqua, 1999; Riley et al., 1989; Robbins& Tanck, 1997; Sperberg &Stabb, 1998; Venable, Carlson, & Wilson, 2001;Zeman, Shipman, & Suveg, 2002). Yet, despitepervasive concern regarding the negative consequences of anger for both affectedindividualsand society as a whole, there has been notably little attention paid to anger in the fields ofsocialscience, including the area of educational psychology. This neglect of research on anger may bedue to the fact that emotions (e.g., anger) are too etherealand complex to study empirically ascompared to behaviors (e.g., violent behavior) (Underwood, 2003). Since the mid 1990s,however, research on anger in the field of social science has begun to make progress (Lemerise& Dodge, 2008), providing the evidence of associationsbetween anger and various negative2outcomes, including bullyingbehavior (Bosworth, Espelage, & Simon, 1999), ganginvolvement(Lemus & Johnson, 2008),substance abuse (Field, 2002; Ryan, Miller-Loessi, & Nieri,2007),low academicperformance (Field, 2002), date aggression (Kinsfogel& Grych, 2004), and peerrejection (Hubbard, 2001). Responding tothe prevalent concern regarding the negativeconsequences of anger, the present studyconsidered anger as a critical element of social-emotional functioning and sought to identifyfactors contributing to the negative impact of angeramong adolescents as predicted froman attachment perspective.From the perspective of attachment theory, Bowiby (1973)claimed that initialattachments to caregivers provide the foundation for theindividual to create internal workingmodels about how relationshipsoperate. These are internalized beliefs or expectations aboutoneself and others in relationships. Bowiby further arguedthat these internal working models canplay a critical role in the experience ofanger. He hypothesized that fearing or/and resistingabandonment or rejection would result in anindividual with an insecure style of attachment whocould become suspicious and hostile, sometimes reactingwith anger whenever the psychologicaldistance from her or his attachment figure increases. Empiricalsupport for this hypothesized linkbetween attachment and anger has been provided (Calamari & Pini, 2003; Kobak, Cole,FerenzGillies, & Fleming, 1993; Kobak & Sceery, 1988; Meesters & Muris,2002; Mikulincer, 1998;Muris, Meesters, Morren, & Moorman, 2004; Troisi & D’Argenio, 2004).However, thesestudies are limited either in the way they assess attachment or in the waythey evaluate anger.The aim of the present study was to adequately examine the Bowlby’s hypothesis of theattachment-anger relationship empirically.This dissertation begins with a background of attachment theory and relevant researchshowing its associations with anger. This is followed by a review of relevant dimensions of anger3as a critical element ofsocial-emotional functioning.Subsequently, a proposed model ofattachment and anger,and hypotheses are presented alongwith specific research questions.Following the description ofthe problem statement, methodological proceduresare presented,further followed by resultsof the present investigation. Finally, asummary and discussion of thefindings, including educational implicationsand limitations of the present study are presented.4Literature ReviewAttachment TheoryA child is busy constructingworking models of how the physical world may beexpected to behave, how his mother and othersignificant persons may beexpected to behave, how he himselfmay be expected to behave, and how eachinteracts with all the others. Within the frameworkof these working models heevaluates his situation and makes hisplans. (Bowiby, 1969/1997,p.354)Nearly half a century ago, John Bowlby (1967/1997)introduced attachment theory, witha significant focus on infants who areregarded as active participants in the attachment process.Bowiby argued that an infant is actively engagedin maintaining secure proximity to her or hiscaregiver(s). From her or his interactions withthe caregiver, the infant learns whether or not thesystem is working well to maintain proximity to thecaregiver. This sense of security providesthe infant with some perception of controlover her or his situation. A caregiver’s availability andresponsiveness to her or his child determines the child’ssense of overall self-worth and thetrustworthiness of other(s) immediately and in future relationships.Stimulated by Bowlby’s work, Mary Ainsworth and hercolleagues examined differentpatterns of attachment in infants and children (Ainsworth,Bleher, Waters, & Wall, 1978). Tothis end, Ainsworth designed the “strange situation”procedure in order to systematically observeon the basis of infants’ or children’s responses to separationfrom and reunion with caregivers.This structured procedure involved a specified series ofepisodes between infant, mother, and“other”: (a) an experimenter introduces a mother and herinfant or child to a playroom, (b) astranger enters and the mother leaves the room,(c) the mother returns to the room, (d) a second5separation leaves the baby or child completely alone, and (e) finally, the stranger and then themother return to the room. After examining hours of the videotaped infants’ or children’sreactions to these episodes, Ainsworth identified three attachment styles: (a) secure, (b)ambivalent/resistant, and (c) avoidant.Children who experienced their primary caregivers as consistently available andresponsive to their signals of distress were identified as having a fundamentally secure pattern ofattachment. ‘When distressed, a secure infant or child actively seeks contact with her or hismother. In contrast, children who experienced their caregiver as inconsistently available andinappropriately responsive were identified as having an ambivalent pattern of attachment.Anambivalent infant or child cannot be certain of her or his caregiver’s availabilityorresponsiveness, and therefore the infant continues to react with expressions ofboth attachmentand anger. Children who experienced their parents as consistently rejectingof their needs wereidentified as having an avoidant pattern of attachment. An avoidant infantor child showsavoidance and detachment, even in the presence of her or his caregiver,presumably for fear ofrejection and punishment. Following Ainsworth’s initial effort to find waysof measuringattachment quality, Main and Solomon (1990) revised Ainsworth’sstudy and proposed theaddition of a fourth attachment style, “disorganized attachment”.A disorganized infant or childdisplays a combination of the ambivalent and the avoidant patternswhen reunited with her or hismother after a short separation (i.e., “cannot classify” codingcategory).Internal working models and traditions ofattachment research.Bowlby (1969/1997,1973) believed that an infant’s feelings of security contributed to healthylater development,whereas infants with insecure attachment were predictedto have less healthy developmentaloutcomes. To provide an explanation for these potential effectsof early attachment styles on6later development, Bowiby proposedthe concept of “internal working models.” For example, if acaregiver consistently gives help andcomfort when needed, the child will develop a workingmodel of the attachment figure asloving and responsive, and of herself or himself as a personworthy of such support. Conversely, if anattachment figure frequently rejects or ignores thechild’s bids for comfort in stressful situations,the child may develop not only an internalworking model of the caregiver as rejectingor unresponsive but also one of herself or himself asnot worthy of help and comfort (Bowlby, 1973).Bretherton (1985) later elaborated on these models, expoundingon the conceptualunderstanding of the idea with her extensively constructive reviewof Bowiby’ s attachmenttheory. Specifically, Bretherton extended Bowlby’shypothesis that an infant’s continuinginteraction with a caregiver would foster the development of an internalworking model in aninfant’s sense of self and other, arguing that, once established, this internalworking model wouldbe stable over time and become an apparatus utilized to interpret events and to determine futureactions. And, indeed, the stability of attachment has been empirically documented inseveralstudies (Gloger-Tippelt, Gomille, Koenig, & Vetter, 2002; Hamilton,2000; Main & Cassidy,1988; Wartner, Grossmann, Fremmer-Bombik,& Suess, 1994; Waters, Merrick, Treboux,Crowell, & Albersheim, 2000). From infancy to middle childhood, for example, 82% - 85% ofattachment styles have been found to be remain the same (Gloger-Tippelt et al., 2002; Main &Cassidy, 1988; Wartner et a!., 1994). Even across the longer period between infancy andadolescence, the stability of attachment styles has been shown to be high. For example, Hamilton(2000) reported a stability rate of 77% from the age of 12 months to the age of 17 years andWaters and colleagues (2000) reported a stability rate of 72% from the age of 12 months to the7age of 21 years. This notion led to further investigations exploringadults’ representations ofchildhood attachment relationships.Main and her colleagues (Hesse, 1999; Main, Kaplan, & Cassidy, 1985) developed theAdult Attachment Interview (AAI) which is a semi-structured interview focusing on attachment-relevant experiences in childhood. The AAI focuses on the dynamics of internal working modelsthat are revealed by the way a person talks about childhood relationships. Based on theseinterviews, individuals can be classified into four distinct attachment groups: (a)secure/autonomous, (b) dismissing, (c) preoccupied, and (d) unresolved/disorganized (i.e.,“cannot classify” interview coding category). These were designed to parallel the four childhoodattachment patterns described earlier: (a) secure, (b) avoidant, (c) ambivalent, and (d)disorganized, respectively.The AAI is also intended to predict the quality of the caregiver’s interaction with her orhis own child and the security of the child’s attachment, as indicated by the Ainsworth StrangeSituation. van IJzendoorn (1995) provided supportive evidence for this link based on his metaanalytic examination, including 22 studies (i.e., 14 studies for study 1, 8 studies for study 2),comparing secure versus insecure representations. Specifically, secure attachment during earlychildhood, as assessed through this retrospective interview, was found to be associated withresponsiveness to their children and secure attachment of their children. Retrospective reports ofinsecure attachments were more likely to be associated with less responsiveness to their childrenand insecure attachment of their children.With an independent research tradition, Hazan and Shaver (1987) began their studies onadolescent and adult romantic attachment, conceptualizing that orientations to romantic or closerelationships might be an outgrowth of previous attachment experiences with parents. Adopting8Ainsworth’ s original three patterns ofchildhood attachment: secure, avoidant, andanxious/ambivalent, Hazan and Shaverdeveloped a self-report questionnaire comprisingparagraph-long descriptions of each ofthe three attachment patterns to assess working modelsofattachment in adulthood. Respondents are asked tochoose a paragraph that describes best theirpattern of attachment to romanticrelationships.Bartholomew (1990) reviewed the attachmentresearch in adolescence and adulthood inboth of these traditions, onefocused on adults’ representations of their childhood relationshipswith parents (Hesse, 1999; Main etal., 1985) and the other focused on romantic or closerelationships (Hazan & Shaver, 1987), andshe concluded with an expanded model of attachmentin adolescence and adulthood (Bartholomew, 1990;Bartholomew & Horowitz, 1991). Throughher careful review of both traditions, Bartholomewnoted two distinct forms of avoidantattachment: dismissing-avoidance characterized with “adefensive maintenance of self-sufficiency and dismissal of attachment needs”andfearful-avoidance characterized with “aconscious fear of anticipated rejection by others”(Bartholomew & Shaver, 1998,p.27), whichwere previously overlooked by researchers and authors.Bartholomew (1990) pointed out that theinterview method by the AAI tended to primarily identifyindividuals who deny attachmentneeds whereas the self-report method by Hazanand Shaver tended to primarily identifyindividuals who fear intimacy although the two approaches definitely identify overlappingavoidant groups. Until Bartholomew reviewed both traditions together, the distinct avoidantattachment styles were not of interest.To systematically address these issues described above, subsequently Bartholomew(Bartholomew, 1990; Bartholomew & Horowitz, 1991) proposed a 2 by 2, or four-group modelof attachment in adolescence and adulthood, returningto Bowlby’s (1973) notion of internal9working models of self andothers. Based on the viewpoint of attachment continuity (i.e.,thecontinuity of an attachment style thought tobe maintained over time), Bartholomew proposedthat an individual’s thoughts,feelings, and behaviors in current close relationships are governedby attachment to their primary caregiversduring childhood. She further argued that models ofself could be dichotomized aspositive (the self is seen as worthy of love and attention) ornegative (the self is seen as unworthy).Similarly, models of others could be dichotomized aspositive (others are seen as available and caring) ornegative (others are seen as unreliable orrejecting). The working models of self and othersjointly define four attachment styles —“secure”, “preoccupied”, “dismissing”, and “fearful”.As indicated earlier, three of these styles —secure, preoccupied, and dismissing— conceptually correspond to the AAI classifications:secure/autonomous, dismissing, and preoccupied,respectively. And three of these styles —secure, preoccupied, and fearful — correspond conceptuallyto Hazan and Shaver’s secure,anxious/ambivalent, and avoidant categories, respectively.The correspondence between Bartholomew’s measure and the AAI, and thecorrespondence between Bartholomew’s and the Hazan and Shaver’s measurehave also beenempirically supported. Specifically, a chi-square study with 30 bereaved women showed that theclassifications obtained from the two measure, Bartholomew’s categories and the AAI’sclassifications, were significantly associated,x2(6)= 24.80,p< .001 (Batholomew & Shaver,1998). Another study (Brennan, Shaver, & Tobey, 1991) of840 college students indicated thatthe classifications obtained from the Bartholomew’s and the Hazan and Shaver’s measures weresignificantly related,x2(6)= 370.31,p< .001.According to Bartholomew (Bartholomew, 1990; Bartholomew & Horowitz, 1991), asecure individual has a sense of love-worthiness and an expectation that other people are10generally accepting and responsive. A preoccupiedindividual possesses a sense of unworthinesscombined with a positive evaluation of others thatwould lead the individual to strive for self-acceptance by gaining the acceptance ofvalued others. A dismissing individual has a sense ofworthiness combined with a negative dispositiontoward others. This individual protects herselfor himself against disappointment by avoidingclose relationships and maintaining a sense ofindependence and invulnerability. Finally, a fearful individual has asense of unworthinesscombined with an expectation that other people will benegatively disposed (i.e., untrustworthyand rejecting). By avoiding close involvementwith others, this individual protects herself orhimself against anticipated rejection by others. Figure 1 provides a useful visual overview of theBartholomew’s conceptual model.wI—F-a)—J0a-MODEL OF SELF(Dependence/Anxiety)Positive Negative(Low) (High)SECURE PREOCCUPIEDComfortable with Preoccupied withintimacy and relationshipsautonomyDISMISSING FEARFULDismissing of Fearful of intimacyintimacy Socially avoidantCounter-dependenta)•00—0G)(13 )zFigure 1. Bartholomew’s (1990) model of adult attachment11To test this model, Bartholomew (Bartholomew & Horowitz, 1991) constructed ameasure of attachment style in adolescence and adulthood which reflects thebasic styles ofattachment that developmental researchers had observed in infants andchildren. Her measure iscomposed of three sub-measures: a self-report measure of experiences in close relationshipsingeneral by revising the Hazan and Shaver’s measure, and two interviews— one on childhoodrelationships along the lines df the AAI and the other on peer relationships(friendships andromantic relationships (Bartholomew & Horowitz, 1991). All threesub-measures rest on thefour-prototype attachment model — secure, preoccupied, dismissing, andfearful. Subsequentresearch (Bartholomew & Horowitz, 1991; Bartholomew& Shaver, 1998) confirmedBartholomew’s initial proposition, demonstrating that individuals’ experiencesin their currentclose relationships as well as their representations of childhoodrelationships with their parentswere consistent with their reported attachment style. That is,individuals who were identified assecure in their current close relationships (friendships and romanticrelationships) were morelikely to be identified as secure in their relationshipswith their parents. Preoccupied individualsin their current close relationships were more likely to bepreoccupied in their relationships withtheir parents, dismissing individuals in close relationshipswere more likely to be dismissing intheir relationships with their parents, and finally fearfulindividuals in current close relationshipswere more likely to be fearful in their relationships withtheir parents.In summary, derived from the Bowlby’s idea of internalworking models of attachment,two independent traditions of attachment research inadolescence and adulthood emerged:one byMain and colleagues (1985) and the other by Hazanand Shaver (1987). By carefully reviewingthe attachment research in both traditions, Bartholomew(1990) systematized Bowlby ‘S internal12working models of self and others, proposing a four-prototype model of attachment inadolescence and adulthood defined in terms of positivity of self and positivity of others.Following this influential revision of attachment model by Bartholomew, some attachmentresearchers (e.g., Brennan, Clark, & Shaver, 1998; Moretti, McKay, & Holland, 2000) have beencontinuing to improve the measurement of attachment in adolescence and adulthood. In the nextsection, important dimensions underlying attachment, which are currently suggested whenassessing attachment in adolescence and adulthood, are discussed.Measuring attachment: Underlying dimensions ofattachment. Grounded in Bowlby’sattachment theory, as reviewed in the preceding sections, researchers have created measures toassess attachment in different developmental stages. Following Ainsworth’ s (1978) study,attachment patterns in infancy and childhood have been primarily measured by observationaltechniques. Adopting the AAI, representations of childhood experiences with parents have beenoften assessed using interview methods (Bartholomew& Horowitz, 1991; Main et al., 1985) foradult populations. Theoretical models of attachment have been refined and still continue to beimproved in consideration of effective assessment of attachment. Thanks to Bartholomew’scritical revision of theoretical models of attachment (Bartholomew, 1990; Bartholomew &Horowitz, 1991), as described earlier, we are now able to understand attachment systematicallyin the working models of self and others.Recent theory suggests that a dimensional approach to understanding attachment (e.g.,continuums of anxiety and avoidance) as opposed to a typological approach (e.g., secure,preoccupied, dismissing, fearful) may be a better way to define and measure attachmentrelationships (Brennan et al., 1998, Fraley & Waller, 1998). Encouraged by Bartholomew’swork, further refinement of underlying the structure of attachment has been made by Brenann13and colleagues (Brenann et al., 1998).Brennan et a!. (1998) suggest that the most establishedtypologies of attachment be recast as a two-dimensionalmodel: anxiety and avoidance.Specifically, Brennan et al. claim thatthe Ainsworth’s three major attachment patterns could beconceptualized as regions in the dimensions of anxiety(i.e., crying, failing to explore confidentlyin the absence of mother, and angryprotest directed at mother during reunions after what wasprobably experienced as abandonment) andavoidance (i.e., discomfort with closeness anddependency). In fact, Ainsworth and her colleagues indicated their acknowledgement of the two-underlying dimensions of attachment in their book (Ainsworth et al., 1978), although theydidnot specifically “name” them, such as anxietyand avoidance. The Main and Solomon’s (1990)now familiar model of infant-attachment styles could also be conceptualized as a product of thetwo underlying dimensions of anxiety and avoidance (see Figure 2).Importantly, Bartholomew’s model of attachment could also be explained in the two-dimensional structure of anxiety and avoidance. Indeed, Bartholomew (Bartholomew, 1990;Bartholomew & Horowitz, 1991) has theorized that her two dimensions (“model of self’ and“model of other”) are also conceptualized in terms of social response styles, that is,“dependence” (or “anxiety”) on the horizontal axis and “avoidance” of intimacy on the verticalaxis (see labels in parentheses in Figure 1). The degree to which the self is viewed as unworthyof love and support (i.e., dependence/anxiety) or significant others are viewed as rejecting orunavailable (i.e., avoidance) determines one’s expectations and behaviors in close relationships(Bartholomew, 1990; Bartholomew & Horowitz, 1991). In other words, a negative model of selfis closely associated with anxiety about abandonment, and a negative model of others is closelyassociated with avoidant behavior (Bartholomew & Shaver, 1998; Brennan et a!., 1998).14Along the lines of thetwo-dimensional (i.e., anxiety and avoidance) approach, Brenannetal. (1998) conducted alarge-sample study, seeking to produce aconceptually as well aspsychometrically valid self-reportattachment measure in romantic relationships.Specifically, byreviewing the attachment literature and extantmeasures of attachment, including some fromunpublished conference papers,Brenann et al. (1998) compiled 60 attachment subscalesand 323relevant items of attachment. After administeringthe 323-item survey questionnaire to 1,086undergraduate students, Brenann etal. conducted a factor analysis to identify underlying factors.Results of factor analysis yielded two essentiallyindependent factors that corresponded to theanxiety and avoidance dimensions. Out ofthe 323 items of the questionnaire, Brenann et al.further developed two refined 18-item scales: one to measure thedimension of anxiety and theother to measure avoidance (i.e., Experiences in CloseRelationships questionnaire; ECR,Brenann, Clark, & Shaver, 1998). Each of the two scales hasdemonstrated high reliability(Cronbach’s alpha = .89 for anxiety, 91 for avoidance;Brenann et al., 1998). Construct validityof these scales are supported by evidence that the anxiety scale is highly correlated with otherscales measuring anxiety and preoccupation with attachment and fear of rejection, and theavoidance scale is highly correlated with other scales measuring avoidance and discomfort withcloseness (Brenann et al., 1998). In addition, the convergence between the ECR measure and theBartholomew’s self-report measure (i.e., the four-clustered categories: secure, dismissing,preoccupied, and fearful) were examined, by clustering participants into four groups. Participantswho scored low on both anxiety and avoidance scales were identified as secure individuals.Participants who scored low on anxiety and high on avoidance were clustered to the dismissinggroup. Those who scored high on anxiety and low on avoidance were identified preoccupiedindividuals, while those who scored high on both anxiety and avoidance were categorized as15fearful individuals. A chi-square test between the twoassessments (i.e., the ECR and theBartholomew’s) was highly significant, indicating considerable similarity between the twomeasurement schemes (Brenatm et al., 1998).wC.)z0>ANXIETYFigure 2. A diagram of anxiety and avoidance in relation to Main and Solomon’s(1990) infant attachment types (reproduced from Brennan, Clark, & Shaver, 1998).Although the ECR measure developed by Brenann and her colleagues were originallydesigned to assess elements of adult romantic attachment, it is possible to expect that much ofthe item content, with some modifications, could be used for measuring adolescents’ attachmentto their caregivers (Moretti et al., 2000) for the following reasons. First, the content of themeasure was thoroughly constructed with the two dimensions, anxiety and avoidance, whichwere derived from extant measures of attachment in romantic relationships grounded inBowlby’s and Ainsworth’s attachment theory originating from child-mother relationships. Asdescribed earlier in this section, Ainsworth and her colleagues (Ainsworth et al., 1978), indeed,pSECURE ANXIOUS-AMBIVALENTDISORIENTED!AVOIDANTDISORGANIZED16had already identifiedthe two dimensions (i.e., anxiety and avoidance)underlying children’sattachment to their mothers in their workalthough they did not call the two dimensions with thespecific names. Second, theECR measure has been found to be conceptually aswell asempirically associated with Bartholomew’sattachment model which was carefully constructed,returning to the Bowlby’s originalinternal working models of self and others. It is also importantto remember that the correspondencebetween the Bartholomew’s measure and the AAI, which isdesigned to tap representations of childhood attachmentwith caregivers, has been empiricallysupported as shown earlier.Responding to limitations of available adolescent attachment measures and a need tocreate one for adolescent populations, Moretti and her colleagues adapted the ECRquestionnaireto develop an adolescent-parent attachment measure, the Comprehensive Adolescent-ParentAttachment Inventory (CAPAI; Moretti et al., 2000). The new measure, consisting the two 18-item subscales of anxiety and avoidance, was created with minor modifications of the ECRquestionnaire, adjusted for reading level and appropriateness of age and context (for example, “Ifeel comfortable sharing my private thoughts and feelings with my partner” was revised to “Ifeel comfortable sharing my private thoughts and feelings with my parent”). A comprehensiveexamination of the psychometric properties of the inventory (i.e., the CAPAI) has beenconducted with a clinical sample of adolescents (age range: 11 year to 17 year-old) to investigatethe reliability and validity of the measure (Steiger, 2003). Results indicated good reliability ofthe measure, a = .89 for anxiety, .91 for avoidance. Analyses of structural validity producedclear two-factor solutions. Further, analyses of convergent validity revealed that each of thesetwo dimensions was correlated with targeted variables. Given that the CAPAI taps the twodimensional model and given statistical support for the validity of the measure, this self-report17measure (i.e., the CAPAI) was used in the present study to assess adolescents’ attachment totheir caregivers in terms of the two dimensions, anxiety and avoidance.In summary, the two-dimensional approach of understanding attachment may have theadvantage of being derived from nearly every other extant attachment measure in closerelationships, including the Bartholomew’s measure, as well as capturing the essence ofBowlby’s and Ainsworth’s attachment theory. In addition, a multi-item dimensional approach asopposed to a typological approach is recommended for statistical reasons. Frarely and Wailer(1998) contend that when classifying people on the basis of their scores, you are necessarilydecreasing the precision of measurement and lowering the statistical power.Of interest in the present study was an examination of attachment precursors of anger. Asinitially proposed by Bowlby (1973), child-parent attachment has implications for howindividuals experience emotions such as anger. The hypothesized link between attachment andanger is addressed in the next section.Anger and an Attachment PerspectiveBowlby (1973) contended that attachment working models play a critical role in theexperience of anger. When experiences lead to the expectation that caregivers will be loving andresponsive, children develop a secure attachment style; that is, they acquire a model of the self asloved and valued and a model of the other as warm and loving. In contrast, when children haveexperiences that lead them to expect caregivers to be rejecting and unreliable, they are likely todevelop an insecure attachment style. These children hold a model of the self as unloved andrejected and/or a model of the others as unloving and rejecting. It has been suggested that aninsecure attachment style contributes to the unhealthy socio-emotional development includinganger (Bowlby, 1969/1997, 1973, 1988).18Bowiby’s original hypothesiswas that dysfunctional anger is a predictablecorrelate ofinsecure attachment. According toBowiby, anger becomes dysfunctionalwhen an individualbecomes so intensely and/orpersistently angry, crossing “the threshold of intensity” (Bowiby,1973,p.249), although it is unclear whathis “threshold” might be. According to Bowiby (1973),the intense anger is initiallydirected toward an attachment figure as aresult of being unloved,rejected, and/or neglected by the attachmentfigure, and then the tendency for anger becomesrepressed and then directed at others. Bowiby (1988)argued that the state of long-term andcommitted relationships, includingrelationships with parents, has a great impact on anindividual’s emotional life, “the underlying tone of how theindividual feels”(p.80). If therelationship goes well, a sense of security isestablished; if the relationship is threatened and thethreats remain consistent, distorted emotionalresponses such as intense levels of anger aredeveloped (Bowlby, 1973, 1988). If an individual’sattachment figure actively rejects her or him,the individual is likely to develop a pattern of responses in which avoidance of the attachmentfigure competes with the individual’s desire for proximity and care, and in which angry feelingsand behavior are apt to become prominent (Bowiby, 1988).Bowlby (1973) stressed that threats of being rejected and abandoned by the attachmentfigure are especially likely to contribute to experience of anger, often of intense degree. Inparticular, repeated threats of abandonment and rejection are expected to lead to the experienceof furiously angry feelings, and this anger, which is used to dissuade the attachment figure fromcarrying out the threat, can become dysfunctional (Bowlby, 1973, 1988).To date, there are seven studies that have addressed Bowiby’s hypothesis regarding angerand attachment. Using the self-report measure developed by Hazan and Shaver (1987) forassessing romantic attachment (i.e., secure, anxious/ambivalent, and avoidant), four studies19(Calamari & Pini, 2003; Meesters & Muris, 2002; Mikulincer, 1998; Muris et al., 2004) havedemonstrated that attachment style is a predictor of anger. Among these, Mikulincer (1998)studied a sample of university students in Israel and found that avoidant individuals reportedhigher levels of hostility than secure and anxious/ambivalent individuals. In addition, Mikulincerreported that anxious/ambivalent individuals scored higher in anger arousal and anger-in(internalized anger) expression and scored lower in anger control than secure and avoidantindividuals. In this study, a self-report measure was used to assess relevant dimensions of angerexpression (i.e., anger-in and anger-out), hostility, anger arousal, anda semistructured interviewwas used to measure anger control.In the Netherlands, two studies (Meesters & Muris, 2002; Muris et al., 2004)of adultsand adolescents, utilizing the Hazan and Shaver’s romantic attachment measure,have shown thatinsecure attachment was associated with greater anger. Meesters and Muris useda self-reportmeasure to assess general anger and hostility. Given the relativelysmall number ofanxious/ambivalent and avoidant adults, the groups of the anxious/ambivalent andavoidantadults were combined as a single insecure-attachment group in this study. Resultsof the studyindicated that insecurely attached adults reported higher levels ofanger and hostility. In anotherstudy, Muris and colleagues (Muris et al., 2004) examined the levels of anger, hostility,and traitanger, reported by secure, anxious/ambivalent, and avoidant attachment groupsbased on theHazan and Shaver’s romantic/close-relationship attachment scale among secondaryschoolstudents. Results of this study revealed that anxiously/ambivalentlyand avoidantly attachedadolescents displayed greater levels of anger and hostility than securely attachedadolescents, butno significant differences in anger and hostility levels werefound betweenanxiously/ambivalently and avoidantly attached groups. Further,Muris and colleagues found that20anxiously/ambivalently and avoidantly attachedadolescents reported higher levels of trait angerthan securely attached adolescents.With female college students inItaly, Calamari and Pini (2003) demonstrated that theavoidant attachment style as assessed by theHazan and Shaver’s romantic-attachment measurewas positively related to anger-in expressionas assessed by a self-report measure.Besides these four studies using the Hazan andShaver’s attachment measure inromantic/close relationships, one study (Troisi & D’Argenio, 2004) inItaly explored theattachment-anger link by utilizing the Bartholomew’s four-groupattachment (i.e., secure,preoccupied, dismissing, and fearful) measure. In their studyof male adults with clinicallydepressive symptoms, Troisi and D’Argenio found that individuals with either thepreoccupiedor the fearful style of attachment reported significantlygreater levels of trait anger as assessed bya self-report measure than those with either the secure or thedismissing style of attachment. Nosignificant differences between preoccupied and fearful individuals and between secure anddismissing individuals were found.Taken together, these five studies (Calamari & Pini, 2003; Meesters & Muris, 2002;Mikulincer, 1998; Muris et al., 2004; Troisi & D’Argenio, 2004) all demonstrated significantassociations between attachment styles and various components of anger. Across the studies,individuals with insecure attachment reported higher levels of anger. However, these studies donot provide direct support for Bowlby’s hypothesis which was based on attachment to parents,not attachment in romantic or close relationships.Two studies (Kobak et al., 1993; Kobak & Sceery, 1988) have explored the relationshipbetween attachment to parents and anger, using the Adult Attachment Interview (AAI; Main etal., 1985), which was designed to assess representations of childhood attachment with parents21(please see the previous sectionsof this paper for details of the AAI). In astudy of collegestudents, Kobak and Sceery (1988)assessed participants’ level of hostility, by askingtheir peersto indicate the degree towhich the subject could be characterized as displayinghostile behavior(e.g., “Has hostility towardothers,” “Expresses hostile feelings directly”). Resultsof this studyindicated that individuals with dismissingattachment were rated higher on hostility by peersthansecure and preoccupied individuals. Nodifferences were found between individuals with secureattachment and those with preoccupiedattachment. Among adolescents, Kobak and colleagues(Kobak et al., 1993) found that insecureadolescents displayed more anger during interactionswith mothers than did secure adolescents. An observationmethod was used to assessparticipants’ anger through their verbal communications(e.g., levels of contempt), nonverbalbehaviors (e.g., sighing), and overt attackingbehaviors (e.g., raising voice levels), during theirinteractions with their mothers.These two studies by Kobak and colleagues are the most relevant to providing empiricalevidence supporting Bowlby’s original hypothesis. The present study extended these studies bylooking at current attachment with parents among adolescents. Research on ongoing attachmentin adolescence has been ignored. This ignorance is seemingly dueto the way attachment researchemerged and developed, moving from examination of attachment in infancy to attachment inadult-romantic relationships, skipping the period of adolescence. Indeed, some researchers(Allen, 2008; Allen & Land, 1999; Thompson, 1997, 1999) have suggested the importance ofassessing ongoing attachment to parents in adolescence. Extending previous research, the presentstudy examined the link between attachment to parents in adolescence with interest in identifyingimportant components of anger to examine Bowlby’s original hypothesis. Previous studiesinvestigating the anger-attachment link have examined a number of different dimensions of22anger, including peer perceptions of hostile behavior (Kobek & Sceery, 1988),observations ofhostile and angry behavior (Kobak et al., 1993), self-reports of hostility (Meesters & Muris,2002; Mikulincer, 1998; Muris et al., 2004;), self-reports of proneness (Calamari & Pini, 2003;Troisi & D’Argenio, 2004), and self-reports of anger expression (Calamari & Pini, 2003;Mikulincer, 1998). Returning to Bowiby’s original hypothesis, the present study examined thelinks between current parental attachment and reported intensity of anger. In addition, thepresent study explored the links between current parental attachment and another potentiallyimportant dimension of anger, anger expression, given research (described previously)demonstrating the critical impact of anger expression on individual’s health. In the followingsection, these dimensions (i.e., intensity and expression of anger) as well as impact of anger arefurther discussed.Dimensions and Impact ofAnger to be ConsideredResearch especially in the clinical and health fields has documented significantassociations between anger and health problems. In particular, the relationship between angerand depression has rich history in psychodynamic theory which holds that anger is a response toincreased tension, discomfort, or frustration and serves as a means of release from this tension(Freud, 19 17/1963; Rubin, 1969; Singer, 1995). However, anger is often perceived as a negativeemotion and/or is blocked or suppressed for a variety of reasons. Defense mechanisms such asdenial, projection, displacement, and rationalization are used by some to suppress anger.According to the psychodynamic view, suppression of anger can result in physical andpsychological conditions, including depression. Consistent with the tenets of psychodynamictheory, several studies have demonstrated a positive relationship between inwardly directedanger and depression (Blumberg & Godwin, 1987; Bridewell & Chang, 1997; Cautin &23Overholser, 2001; Chaplin, 2006; Clayet al, 1993; Golman & Haaga, 1995; Kopper & Epperson,1996; Newman et al., 1999;Riley et al., 1989; Robbins & Tanck, 1997; Sperberg& Stabb, 1998;Venable et a!., 2001; Zeman et al., 2002).That is, the more individuals suppress their expressionof felt anger, the greater theirrisk for depression. Recently, the association of suppressed orinwardly directed anger with other health problems havebeen reported. Female adolescentsreporting greater levels of eating disorder symptomswere more likely to inhibit anger feelings(Zaitsoff, Geller, & Srikameswaren, 2002). As well, increasedlevels of suppressed anger werefound in college students with obsessive-compulsivedisorder (Whiteside & Abramowitz, 2005).Furthermore, adolescent psychiatric inpatients with a tendency to internalize angerwere at riskfor suicide attempts (Cautin & Overholser, 2001).Outwardly directed anger, as opposed to suppressed or inwardly directed anger,is alsoassociated with health problems. Research (Bridewell &Chang, 1997; Cautin & Overholser,2001; Kopper & Epperson, 1996; Riley et a!., 1989; Sperberg &Stabb, 1998) has demonstrated apositive relationship between outwardly directed anger and health problems. Higherlevels ofoutwardly directed anger have been found among individuals with posttraumatic stress disorder(Riley et al., 1989). Adolescents reporting greater levels of externalizinganger were more likelyto abuse alcohol (Cautin & Overholser, 2001). Similar to individuals reporting higher levels ofinwardly directed anger, individuals demonstrating higher levels of outwardly directed angerwere also at risk for depression (Bridewell & Chang, 1997; Kopper & Epperson, 1996; Sperberg& Stabb, 1998).Given the negative impact of greater levels of anger inwardly directed and outwardlydirected on individuals’ health, it becomes important to understand the factors that contribute toboth a boost and reduction in anger expression. Bowlby’s initial hypothesis concerned the24relationship between attachment to parents and intensityof anger. This hypothesized link wasaddressed in the present study in a sample of adolescents. In addition, thepresent study extendedthe examination of the Bowlby’shypothesis by looking at potential impact of another dimensionof anger, the expression of anger (i.e., inwardlydirected anger and outwardly directed anger), inaddition to the dimension of anger, intensity of anger felt. Although all individuals mayexperience feelings of anger, the way in which such feelings are expressed may varyconsiderably. Indeed, Spielberger (1999) has pointed out the importance of considering both,intensity and expression of anger, distinguishing these two dimensions, in examining theirassociations with other relevant health problems.Spielberger (1988, 1999) has defined anger as an emotional state consisting of feelingsthat vary in intensity (levels) and expression. Originally, Spielberger developed a self-reportmeasure, the State-Trait Anger Scale, to assess the intensity or frequency of anger experienced(STAS; Spielberger, Jacobs, Russell, & Crane, 1983). Using the STAS, the intensity of angeramong male hypertensive patients with those of a control group of general patients having nohistory of hypertension was compared (Spielberger et al., 1983; Spielberger, Johnson, Russell,Crane, Jacobs, & Worden, 1985). Results showed that the hypertensive group reportedexperiencing more intense anger than the control group. However, the hypertensive individualsappeared to suppress these feelings in interpersonal situations, resulting in less overt aggressivebehavior. For Spielberger, these observations underscored the importance of assessing the extentto which individuals express (or suppress) their anger feelings, in addition to assessing theintensity of anger feelings (Spielberger et al., 1985).Later, Spielberger (1988, 1999) identified two distinguishable dimensions regarding theexpression of anger: (a) anger inwardly suppressed (anger-in) and (b) anger expressed outwardly25toward others (anger-out). Individualshigh in the anger-in dimension may experienceintenseanger, but suppress rather thanexpress these feelings. In contrast, people high on anger-outfrequently display their anger viaaggressive verbal or physical behavior. Speilberger developedthe State-Trait Anger Expression Inventory(STAXI-2) to measure each of these components ofanger. The intensity of angry feelingsis assessed as the disposition to experience anger by thetrait-anger scale of the STAXI-2. The tendencyor frequency of suppressing angry feelings (i.e.,anger-in) is measured by the anger expression-in scale, andthe frequency of expressing angeroutwardly (i.e., anger-out) is assessed by theanger expression-out scale of the STAXI-2.The trait-anger scale, the anger-in scale, and the anger-out scale of theSTAXI-2 wereused in the present study. The reliability and validity of theSTAXI-2 have been demonstrated inprevious research (Spielberg, 1999). Detailedevidence supporting the psychometrical andconceptual structure of the scale is described in the method section.In the theoretical literature, sex differences in anger expression have been contended(Fischer, Smith, Leonard, Fugua, Campbell, & Masters, 1993; Sharkin, 1993). That is, womenare believed to be more likely to mask or suppress their feelings of anger than men. However,empirical studies do not support this hypothesis, providing no evidence of significant sexdifferences in anger expression at least in adult populations (Chaplin, 2006; Kopper & Epperson,1996; Newman et al., 1999). One study (Cox, Stabb, & Hulgus, 2000) examining a child andadolescent population, however, has demonstrated significant sex differences in angerexpression. In their study, with a sample of school children from grades 5-9, Cox and colleaguesfound that girls scored significantly higher on anger-in than boys regardless of age or gradegrouping. Accordingly, it is important to further examine potential sex differences in the presentstudy with a sample of adolescents.26Statement of the ProblemIn light of the theoretical and empirical literature reviewed thusfar, it was not clear thatthe Bowiby’s original hypothesis regarding therelationship between attachment to parents andanger has as yet been adequately tested empirically.Moreover, given recent advances, boththeoretical and methodological, regarding the assessment of both attachment andanger, a morerefined test of the hypothesis was conceivable. Based on the literature extant, a modelillustratingthe relationship between attachment to parents and anger was proposed and provided a basis forthe present investigation.The theoretically hypothesized model is illustrated in Figure 3. Consistent with theBowiby’s original hypotheses, this model predicted that adolescents’ attachment to parentswould influence the intensity of anger felt generally. Further, the present study extendedBowiby’s hypothesis, considering another important component of anger, anger expression, inaddition to the intensity of anger. Of particular interest in the present study was an investigationof how two distinct dimensions of anger, intensity and expression, were related to currentattachment to parents in a sample of adolescents. Even though all individuals might experiencefeelings of anger, the way that such feelings are expressed might differ notably. It was expectedthat high levels of both attachment anxiety and avoidance would contribute to greater levels ofanger feelings (i.e., the intensity of felt anger) which, in turn, would elevate the levels of bothanger-in (i.e., suppressing anger) and anger-out expressions. Furthermore, it was expected thatthere would be a direct effect of attachment anxiety on anger-in expression and a direct effect ofattachment avoidance on both anger-out and anger-in expressions, in addition to the mediatedrelationship between attachment and anger expression through the intensity of anger. Specifichypotheses are further described below, along with research questions in the present study.27Question 1: Isthere a relationship betweenattachment dimensions and angerexperienceand expression? Ifso,how are they related?Hypothesis]: Dimensionsof insecure attachment (i.e.,attachment anxiety and attachmentavoidance) were expected to bepositively related to increases in feelingsorintensity of anger amongadolescents which, in turn, would be associatedwithincreases in anger-in and anger-outexpressions (i.e., mediationalrelationships). This hypothesis wasprimarily based on Bowlby’s originalhypothesis that attachment experience withparents would predict intensity ofanger. Two studies (Kobak et al., 1993; Kobak& Sceery, 1988) have providedsupport for this hypothesis, using the AAI,tapping retrospective perceptions ofearly attachment to parents. Thepresent study extended these studies bylooking at current attachment to parents duringadolescence. Anger expressionwas considered as another important dimensionof anger in the present studybased on the Spielberger’s (1988, 1999)argument that it is important todistinguish anger expression from anger experience/intensity.Given evidenceof the negative impact of both anger-in and anger-outexpression on health(Blumberg & Godwin, 1987; Bridewell & Chang, 1997;Cautin & Overholser,2001; Chaplin, 2006; Clay et al., 2001; Golman & Haaga,1995; Kopper &Epperson, 1996; Newman et al., 1999; Riley et al., 1989; Robbins &Tanck,1997; Sperberg & Stabb, 1998; Venable et al., 2001;Zeman et al., 2002;Zaitsoff et al., 2002; Whiteside & Abramowitz, 2005), the importanceofmeasuring the extent to which individuals suppress or express theirangerfeelings has become increasingly apparent, in addition to assessingthe intensity28of anger feelings. Anger intensity,defined as the experience of anger as anemotional state by Spielberger (1999), wastreated as a mediator precedinganger expression in the present study. Thishypothesized mediation was derivedfrom the emotion literature suggesting thatemotional expressions are themanifestations of internal emotional states (Lewis, 2008; Lewis &Michalson,1983). Accordingly, in the present study, felt anger was considered anecessaryprerequisite for the expression of anger, a hypothesis tested by a meditationalmodel. In the present study, it was expected that the increased levels of angryfeelings (i.e., intensity of anger) would be predicted by attachment anxiety andthat attachment avoidance would contribute to increased levels of both anger-inand anger-out expressions.In addition, an exploratory examination of attachment anxiety xattachment avoidance interaction effects on anger experience was conducted.Considering the employment of the two-dimensional scale of attachment (i.e.,anxiety and avoidance) in the present study, it was expected that, in addition tofinding a main effect of each attachment dimension (i.e., anxious andavoidant), an interaction between the two dimensions was also expected.Furthermore, both attachment anxiety and attachment avoidance wereexpected to be positively associated with anger-in. However, anger-out wasexpected to only be linked to attachment avoidance. This hypothesis was drawnfrom the findings of previous studies demonstrating that suppressed anger oranger-in is associated with both anxious/ambivalent attachment in romanticrelationships (Mikulincer, 1998) and avoidant attachment (Calamari & Pini,292003), and the finding by Kobakand Sceery (1988) regarding theassociationbetween dismissing (or avoidant)attachment to parents and anger rated bypeers, implying a type of anger-outexpression.Question 2: Are there sexdifferences in the attachment-anger link?Hypothesis 2: Given the evidence of one study(Cox et a!., 2000) with a sample of studentsingrades 5-9 that girls scored higher onanger-in expression than boys, sexdifferences were considered in the presentstudy. Of interest was whether thepresent study would replicate the Cox et al.’sfinding (i.e., girls score higher onanger-in than boys) and whether the links between attachment toparents andanger intensity and expression might vary for male and female adolescentsalthough no specific hypothesis were made in this regard, given thelimitedevidence to date regarding sex differences in the hypothesized relationshipbetween attachment and anger.Question 3: Does the relationship betweenattachment and anger differ between attachmentfigures, mother andfather?Hypothesis 3: Previous studies (Kobak et al., 1993;Kobak & Sceery, 1988) showing thesignificant relationship between attachment and anger did not investigatedifferences between mother and father figures in terms of the attachment-angerlink. The attachment classifications assessed by the AAI in Koback andcolleagues’ studies (Kobak et al., 1993; Kobak & Sceery, 1988) do not informus about which attachment figure was examined. When employing theAAI, theattachment to mother and attachment to father are usually coalesced to producean attachment classification for an individual. However, it may beworthwhile30to note the Bowiby’ (1967/1997)concept of”monotropy” that a child has aninnate need to attach to one primaryattachment figure, usually the mother, withregard to the exploration ofmother-father differences. Even though Bowibynever ruled out the possible presence ofother attachment figures for a child, hedid clearly indicated that there was a primarybond which was much moreimportant than others, usually the child’s natural or biologicalmother (Bowiby,1967/1997, 1988). In most cases, the biological mother who has thegreatestbiological investment in the child could be most influential in the developmentof the child (Bowiby, 1967/1997, Cassidy,2008). If Bowlby’s hypothesis iscorrect, attachment to the mother figure would have greater influence on theexperience of anger than attachment to the father figure in the present study.Accordingly, independent contributions of attachment to both mother andfather figures were examined in the present study.In addition, far less is known about the putative influence of child-fatherattachment, given that the relevant parent in most previous attachment studieshas been solely the mother (Cassidy, 2008; van IJzendoorn & De Wolff, 1997).van IJzendoom and De Wolff (1997) argue that “although we do not want tospeculate about the question of whether the absent father is a social or ascientific construction, ... the dearth of studies on the role of the father ininfants’ development of attachment should unfortunately be considered amatter of fact instead of (sexist) opinion” (p. 604). No specific hypothesis wasproposed concerning differences between mother and father figures, because ofinsufficient prior empirical research linking these variables.31Figure 3. A model of attachment and anger.Note. The arrowed-solid lines denote significant paths hypothesized in the presentstudy,whereas the dashed line denotes a non-significant path hypothesized.32MethodParticipantsStudents in grades 8-12 from four secondaryschools in the southern British Columbia,Canada, were recruited for participationin the present study. Of these students, participantsincluded 776 students (379 boys, 397 girls),ranging in age 13 to 19 years (M= 15.2, SD = 1.58),who had received parentalconsent and who themselves agreed to participate. Theoverallparticipation rate was 78%. Students from a varietyof ethnic backgrounds were included: 53.6%Asian Canadian, 20% European Canadian, 7.5%South-Asian Canadian, 2.4% Middle-EasternCanadian, 1.4% Latino Canadian, 0.8% First Nations, 0.8%AfricanlCaribbean Canadian, 10.2%Mixed, and 3.4% “Other” (non-specified).The distribution of the participants by grade level andsex is presented in Table 1.Table 1Distribution ofParticipants by Grade Level and SexSexBoys (n) Girls (n) TotalGrade 8 113 97 2109 46 46 9210 67 73 14011 91 88 17912 62 93 155Total 379 397 77633ProceduresStudents in eighth- throughtwelfth-grade classrooms were asked to take home a letter totheir parents explaining the purpose and natureof research, acknowledging that students’responses would be considered confidential, and asking parentsfor permission for their son ordaughter to participate in the research. Studentswho received parental permission (see AppendixA-i) and who themselves agreed toparticipate (see Appendix A-2 for the student assent form)were involved in a single group-testing session (30-50minutes) during which a self-reportsurvey was administered in each classroom by at least two trainedproctors (the author andtrained graduate students). Teachers remained in the classroomto oversee student discipline butwere otherwise uninvolved. Prior to the administration of the survey,the study was reviewed andapproved by the Behavioural Research Ethics Board at the University of British Columbia andthe participating school districts (see Appendix B for approval).MeasuresDemographic information. This measure gathers demographic background information,including (a) sex, (b) birth date and age, (c) grade, and (d) ethnic background (see Appendix Ci).Attachment. To assess students’ ongoing attachment to their caregivers, theComprehensive Adolescent-Parent Attachment Inventory (CAPAT; Moretti et al., 2000) was used(see Appendix C-2). As described previously, this attachment measure is a 36-item, self-reportmeasure designed to assess adolescents’ attachment on the basis of the two-dimensionalstructure, attachment anxiety and attachment avoidance. Attachment anxiety refers to thefeelings of insecurity about not feeling close to parents accompanied by a low level of self-34sufficiency, while attachmentavoidance is related to thedevaluing and dismissing needfor aclose relationshipwith a parent (McKay &Moretti, 2001). Participantswere asked to respond toquestions about their relationshipswith their mother and father figureson a 7-point Likert-typescale (1 = disagreestrongly to 7 = agree strongly).An example of an itemincluded in thisinventory is as follows: “Ifeel comfortable dependingon my mother.” Participants’responses toitems in each of the two-dimensionsubscales of the CAPAI (18 itemsfor each dimensionsubscale) were averagedto create an overall (i.e., mean)index for each attachmentdimension,with higher scores reflectinggreater levels of attachmentanxiety or attachment avoidance.Each of the two-dimension subscales,attachment anxiety and avoidance, hasdemonstrated high internal consistencyin a similar age (ages: 11-17) clinical sample(Cronbach’s alpha .89 for anxiety,91 for avoidance; Steiger, 2003). Convergentvalidity forthis measure has been supportedby comparing attachment ratings toconcurrent measures ofpsychopathology such as internalizingsymptoms, externalizing symptoms,anxiety, anddepression, which have beentheoretically linked with each other.In the original CAPAI, respondentsare asked to indicate the parent or caregiver whotheyfeel has “played the most importantpart in raising” them and answer in terms oftheirrelationships with that person. Ofinterest in the present study was an exploration ofwhichattachment figure, mother or father, mighthave a stronger impact on the hypothesizedattachment-anger link. Accordingly,in the present study, the CAPAI survey wasadapted toassess adolescents’ relationshipswith their mothers (or mother figures) andfathers (or fatherfigures) separately.Anger. To measure different facets ofstudents’ anger (i.e., the intensity of anger felt andanger-expression), trait-anger,anger-in, and anger-out scales of a self-reportmeasure, the State-35Trait Anger Expression Inventory 2(STAXI-2; Spielberger, 1999), were used (seeAppendix C-3). The trait-angerscale consisting of 10 items is designed to measure the intensityof angryfeelings as the disposition toexperience anger (e.g., “I am quick-tempered”). The angerexpression-in (i.e., anger-in) scale is an 8-item measurethat assesses the tendency or frequencyof suppressing anger feelings (e.g.,“I keep things in”). The anger expression-out (i.e.,anger-out)scale consisting 8 items measures the frequencyof expressing anger outwardly (e.g., “I expressmy anger”)’. Participants wereasked to rate dimensions of their anger on a 4-point Likert-typescale (1 = almost never to 4 = almost always).Students’ responses to relevant items wereaverage to create overall indices of anger acrossthe intensity of anger felt, anger-in expression,and anger-out expression. Higher scores reflected greaterlevels of anger in each case.The reliability and validity of the STAXI-2 have been demonstrated in previousresearch(Spielberg, 1999). With a sample of college students (age range: 16-19), the internalconsistencyreliabilities of the scales (i.e., trait anger, anger-in and anger-out) are satisfactory: .89for malesand .88 for females in trait anger, .74 for males and .79 for females in anger-in, .78for males and.76 for females in anger-out (Spielberger, 1999). The concurrent validity ofthe trait-anger scaleof the original STAXI (note: items of the STAXI and STAXI-2 for this scale are identical) hasbeen evaluated and supported by evidence that the scale is significantly correlated withotheranger measures, the Buss-Durkee Hostility Inventory (BDHI) andthe hostility scale of theMinnesota Multiphasic Personality Inventory (MMPI) in samples of college students and Navyrecruits (Spielberger, 1999). The convergent and divergent validity of the anger-in and anger-outscales has been supported by comparing the scales to other measures of angerexpressionThese anger scale items have been reproduced by special permission ofthe Publisher, Psychological AssessmentResources, Inc., 16204 North Florida Avenue, Lutz, FL, 33549, from the STAXI-2 by Charles D.Spielberger,Ph.D., Copyright 1979, 1986, 1988, 1995, 1998, 1999, by PsychologicalAssessment Resources, Inc. Furtherreproduction is prohibited without permission from PAR, Inc.36(Spielberger, 1999; Spielberger et al., 1985). Specifically, with a sample of high school students,students were grouped as either “anger-in” or “anger-out” based on the students’ responses tovignettes which described anger-provoking situations developed by Harburg and colleagues(Harburg, Blakelock, & Roeper, 1979). Subsequently, differences in the anger-in and anger-outscores of the original STAXI (note: items of the STAXI and STAXI-2 for these scales areidentical) were evaluated based on the classifications of anger-in and anger-out groups identifiedby the Harburg’ s measure. Results indicated that the anger-in group classified by theHarburgmeasure had significantly higher scores on the anger-in scale of STAXI-2 andsignificantly lowerscores on the STAXI-2 anger-out scale, whereas the anger-out group of the Harburg measure hadsignificantly higher scores on the STAXI-2 anger-out scale and significantly lower scores on theSTAXI-2 anger-in scale. Given the evidence supporting the psychometrical andconceptualstructure of the anger scales, the trait-anger scale, the anger-in scale, and theanger-out scalefrom the STAXI-2 were employed in the present study to assess the relevant dimensionsofanger.37ResultsData Preparation and ScreeningMissing data. There were 15 cases reporting no responses (2% ofthe total sample)2inthe attachment variables (i.e., anxiety and avoidance).Of the 15 cases, one case (girl) had noresponses on mother-figure attachment questions and 14 cases (10 boys,4 girls) reported noresponses on father-figure attachment items. Inthe present study, the following six data sets werecreated to examine the hypotheses of the present study: (a) motherfigure attachment with theentire sample, (b) father figure attachment with the entire sample, (c) mother figure attachmentwith boys only, (d) mother figure attachment with only girls, (e) father figure attachment withonly boys, and (f) father figure attachment with only girls. One case having no responses onmother-figure attachment questions was eliminated from analyses on the mother-figureattachment, and the 14 cases with no responses on father-figure attachment items were droppedfrom analyses on the father-figure attachment. The final sample sizes for the six data sets arepresented in Table 2.2No further missing data values were found in the present data after careful inspection of missing data, including thedetection of possible spontaneous response patterns (e.g., circling the same number of the questionnaire items for allanswers). The low rate of the missing data in this study was obtained as a result of the use of thorough reminders bytrained proctors’ to participants not to miss a question during the survey time.38Table 2Sample Sizesfor Data Analyses____________________________NoteOne case with no responses on mother-figure attachmentitems was dropped from the original entire sample size, 776.14 cases with no responses on father-figure attachmentitems were dropped from the original entire sample size, 776.None of cases was dropped from the original sample size forboys, 379.One case (one girl) with no responses on mother-figureattachment items was dropped from the original sample sizefor girls, 397.10 cases (10 boys) with no responses on father-figureattachment items were dropped from the original sample sizefor boys, 379.4 cases (4 girls) with no responses on father-figureattachment items were dropped from the original sample sizefor girls, 397.Tests ofassumptions. Priority to the examination of the primary hypotheses in thepresent study, normality, linearity, and outliers of all variables were evaluated. Normality of thevariables was assessed through visual examination of histograms and indices of skewness andkurtosis. Values for skewness and kurtosis were considered to indicate a normal distribution ifthey were< 121(Miles & Shevlin, 2001). As seen in Table 3, none of the skewness and kurtosisvalues for the observed variables exceeded the cut-off point (i.e.,121), indicating a normaldistribution for each of the observed variables. Normality of the variables was also visuallyconfirmed with frequency histograms.Data SetMother figure attachment: EntireFather figure attachment: EntireMother figure attachment: BoysFinal samplesize (n)775762379Mother figure attachment: Girls 396Father figure attachment: Boys 369Father figure attachment: Girls 39339Linearity was diagnosed from bivariate scatterplots between pairs of variables. None ofthe plots appeared to suggest a non-linear relationship in any of the cases, supporting theassumption of linearity.Outliers were inspected with standardized scores (i.e., z scores) for univariate outliers andwith Mahalanobis distance statistics for multivariate outliers. Cases with z scores in excessof14.01are potential univariate outliers (Hair, Black, Babin, Anderson, & Tatham, 2006). Therewere no univariate outliers found in the present data sets. With regardto multivariate outliers,Mahalanobis distance statistics atp < .001 were calculated asx2with degrees of freedom equalto the number of variables, in this case, six. Then, any case with a Mahalanobis distance greaterthanx2(6)= 22.46 is a multivariate outlier. Four cases were identified as multivariate outlierswith the data set of entire sample of mother figure attachment (N= 775). Three cases with thedata set of entire sample of father figure attachment (N = 762), three cases with thedata set ofmother figure attachment for only boys (n = 379), one case with the data set of motherfigureattachment for only girls (n = 396), and two cases with the data set of father figureattachmentfor only boys (n = 369), were also identified as multivariate outliers. Subsequently,these caseswere examined, and none of them appeared to be outside of the rangeof expected values andtheir overall patterns of scores across all of the variables made conceptualsense. Hence, none ofthese cases were removed from the analyses.40Table 3Results ofNormality, and DescriptiveStatisticsfor Study VariablesM SD Skewness KurtosisVariableMother figure attachment: Entire (N = 775)Attachment anxiety 2.56 0.84 0.60 0.22Attachment avoidance 3.211.18 0.25 -0.59Intensity of anger 1.970.57 0.69 0.34Anger-in 2.200.54 0.25 -0.37Anger-out 2.020.51 0.58 0.12Father figure attachment: Entire (N = 762)Attachment anxiety 2.59 0.93 0.46 -0.26Attachment avoidance 3.731.32 0.05 -0.74Intensityofanger 1.980.57 0.68 0.33Anger-in 2.20 0.540.25 -0.36Anger-out2.02 0.51 0.59 0.12Mother figure attachment: Boys (n = 379)Attachment anxiety 2.44 0.79 0.60 0.20AttachmentAvoidance 3.331.10 0.19 -0.56Intensity of anger 1.990.57 0.56 0.22Anger-in 2.19 0.550.30 -0.31Anger-out 2.04 0.500.58 0.31Mother figure attachment: Girls (n = 396)Attachment anxiety 2.68 0.87 0.56 0.16Attachment avoidance 3.101.24 0.34 -0.60Intensity of anger 1.95 0.570.82 0.52Anger-in 2.20 0.540.21 0.21Anger-out 2.00 0.510.60 -0.0241Father figure attachment: Boys (n =369)Attachment anxiety 2.450.90 0.55 -0.15Attachment avoidance3.68 1.26 0.05-0.59Intensityof anger2.00 0.58 0.560.20Anger-in2.20 0.55 0.30-0.31Anger-out2.04 0.50 0.59 0.32Father figure attachment: Girls (n = 393)Attachment anxiety 2.710.94 0.38 -0.29Attachment avoidance3.77 1.38 0.04 -0.87Intensityofanger1.96 0.57 0.81 0.50Anger-in2.20 0.54 0.21 -0.40Anger-out2.00 0.51 0.59 -0.03Bivariate correlations were calculated to assess theinterrelations among all variables. Asseen in Table 4, small but significant positive correlationswere observed between attachmentanxiety and attachment avoidance subscales in alldata sets, except the data set of father figureattachment for boys, suggesting the twodistinct dimensions (i.e., attachment anxiety andattachment avoidant) but they are under the sameumbrella of a broad construct — attachment.Small to moderate correlations in the expected directions were found amongthe anger subscales,as shown in Table 4. Overall, these patterns of the correlationswere consistent with theoreticallyexpected relationships.42Table 4Intercorrelations Among Study Variables1 2 3 4 5VariableMother figure attachment: Entire (N = 775)1. Attachment anxiety 1.002. Attachment avoidance .13**1.003. Intensity of anger.24** .24**1.004. Anger-in.21** .30**44**1.005. Anger-out.16**.19.67** .29**1.00Father figure attachment: Entire (N = 762)1. Attachment anxiety 1.002. Attachment avoidance.12**1.003. Intensity of anger.18** .24**1.004. Anger-in.20** .28**45**1.005. Anger-out.12** .67** .30**1.00Mother figure attachment: Boys (n = 379)1. Attachment anxiety 1.002. Attachment avoidance .15**1.003. Intensity of anger.19** .18**1.004. Anger-in.25** .31** .40**1.005. Anger-out.13* .10* .65** .22**1.00Mother figure attachment: Girls (n = 396)1. Attachment anxiety 1.002. Attachment avoidance.14**1.003. Intensity of anger.31** .29**1.004. Anger-in.19** .30** .48**1.005. Anger-out.20** .26** .69** .36**1.0043Father figure attachment: Boys (n = 369)1. Attachment anxiety 1.002. Attachment avoidance .091.003. Intensity of anger.18** .21**1.004. Anger-in.25** .28** .41**1.005. Anger-outj3*.12*1.00Father figure attachment: Girls (n = 393)1. Attachment anxiety 1.002. Attachment avoidance1.003. Intensity of anger.26**1.004. Anger-in.15** .27**49**1.005. Anger-out.11*37**1.00*p<.05.**p<.01.44Assessment of unidimensionality among study variables. Next, an exploratory factoranalysis was conducted, using the Mplus 5.1 program (Muthen & Muthén, 2007), on eachvariable used in the present structural equation modeling analyses, in order to examine thedimensionality of each measurement instrument (i.e., variable). In other words, each variablewas examined to determine whether it was unidimensional. Considering the categorical nature ofthe measurements used in the present study (i.e., Likert-type scales), polychoric correlations andthe Weighted Least Squares (WLS) estimation method were used in the exploratory factoranalyses. Following the Ford et al.’s (Ford, MacCallum, & Tait, 1986) and other authors’(Conway & Huffcut, 2003; Fabrigar, Wegener, MacCallum, & Strahan, 1999; Floyd, Widaman,1995; Gorsuch, 1997) recommendations, the oblique (i.e., promax) rotation was used. Obliquerotation allows factors to be correlated, whereas orthogonal rotation produces factors that arestatistically uncorrelated, which is highly unlikely in real world assessments.To determine whether essential unidimensionality was observed, denoting the presence ofa reasonably dominant common factor along with secondary minor dimensions (Nandakumar,1993; Nandakumar & Ackerman, 2004; Staout, 1987; 1990), the following criteria were utilized:(a) the ratio of the first to the second eigenvalue(>3.0) (Morizot, Ainsworth, & Reise, 2007), (b)factor loadings(>.1301)(Nunnally & Bernstein, 1994), (c) the comparative fit index (CFI; > .90)(Hu & Bentler, 1999)., and (d) the root-mean-square error of approximation (RMSEA; good fit:<.05; .mediocre fit: .08 - .10) (Browne & Cudeck, 1993; Hu & Bentler, 1999). Cases that did notmeet any of these criteria were considered for alteration. Factor analyses were conducted for theentire sample and for boys and girls separately. Accordingly, primary analyses examining thehypothesis of the present study were performed on the following six separate data sets: (a)45mother figure attachmentwith the entire sample, (b) fatherfigure attachment with the entiresample, (c) mother figureattachment with boys only, (d) mother figureattachment with onlygirls, (e) father figureattachment with only boys, and (f) father figureattachment with only girls.As seen in Table 5, results showed that therewere three cases which did not meet any ofthose criteria described above: (a) theanger-in variable in mother figure attachment withonlygirls, (b) the anger-in variable infather figure attachment with only girls, (c) the attachmentanxiety variable in father figure attachmentwith only girls. Although attempts were made toidentify better measurementstructures by removing items which did not obtain factorloadingshigher than .30, none of these attemptsresulted in substantially better unidimensionality.Inaddition, those items attempted to be removed (i.e.,11, X9 and X183 in Table 5) are stilltheoretically and conceptually relevantto the constructs of the variables. Furthermore, based onvisual evaluations with scree plots showing eigenvalues andoverall intercorrelations among thescale items, all of the variables were considered withinacceptable range of unidimensionality(see Appendix D for scree plots and Appendix E foritem intercorrelations). Therefore, noalternations were made to the variables in this study.However, given that the three variableswere marginally within acceptable range of unidimensionality, resultsfor girls should beinterpreted with caution.The first letter of these items indicates the scale it belongs to (i.e., X= attachment anxiety, V = attachmentavoidance, T = intensity of anger, I = anger-in, 0 = anger-Out). The number nextto the letter denotes the itemnumber in the scale.46Table 5Results ofFactor Analysesfor Study VariablesEigenvalue Factor Reliability4ratio loadings CFI RMSEA (ci)Mother figure attachment: Entire (N = 775)Attachment anxiety (item X9)3.05.26*- .76 .86 .14 .84Attachment avoidance 5.87 .47 - .82.95 .17 .93Intensity of anger 3.43 .57 - .85.91 .18 .84Anger-in 2.18 .32 - .77 .84.16 .71Anger-out (item 02) 3.07.27*- .73 .96 .08 .73Father figure attachment: Entire (N = 762)Attachment Anxiety (item X18)3.53.22*- .80 .89 .16 .86Attachment avoidance 6.10 .43 -.84 .95 .21 .94Intensityofanger 3.41 .56- .84.91 .18 .84Anger-in 2.17.33 - .77 .83 .17 .71Anger-out (item 02) 3.06.27*- .68 .96 .08 .73Mother figure attachment: Boys (n = 379)Attachment anxiety3.30 .31 - .71 .89 .13 .83Attachment avoidance 4.87.38 - .83 .94 .17 .92Intensity of anger 3.12 .57- .82 .90 .19 .84Anger-in 2.56 .37 - .76.87 .15 .73Anger-out (item 02) 2.62.20*- .71 .94 .09 .71Mother figure attachment: Girls (n = 396)Attachment anxiety (item X9)3.01.17*- .83 .86 .15 .84Internal consistency of each instrument or variable (i.e., Cronbach’s coefficientaiphas) was presented in Table 4 asa reference only. Although the coefficient alphas have been most commonly usedas the index of unidirnensionalityin previous research, they need to be distinguished from a function of unidimensionality(Hattie, 1985).Unidimensionality can be defined as the existence of one major latent traitunderlying data, and it may notnecessarily be internally consistent (Hattie, 1985).47Attachmentavoidance6.44 .55- .84 .96 .18 .94Intensity of anger 3.79.57 - .86 .93 .17 .85Anger-in (item II) 1.96.24*- .77 .82 .17 .69Anger-out3.52 .34 - .75 .98 .07 .75Father figure attachment: Boys (n = 369)Attachment anxiety (item X18) 4.80.19*- .83 .94 .13 .87Attachment avoidance 4.94.36 - .80 .94 .21 .94Intensity of anger 3.08.57 - .82 .89 .20 .84Anger-in2.53 .39 - .75 .86 .16 .73Anger-out (item 02) 2.64 .21*- .70 .94 .09 .71Father figure attachment: Girls (n = 393)Attachment anxiety (item X18) 2.86.23*- .79 .85 .19 .85Attachment avoidance 7.36 .49 - .87 .96 .19 .95Intensity of anger 3.80 .60 - .86 .93.17 .85Anger-in (item Ii) 1.96.23*- .77 .82 .17 .69Anger-out 3.49 .34 - .75 .97 .07 .70Note.*denotes weak factor loadings (i.e., < .301). Each of the scales/variables with*possessed only oneitem with weak factor loading.General sex differences on study variables. Although sex differences were notanticipated, a series oft tests were conducted to assess general sex differences on each of thepredictor and outcome variables. As shown in Table 6, the tests for attachment anxiety to bothmother and father and attachment avoidance to mother were statistically significant, but therewere no significant differences between boys and girls for attachment avoidance to father,intensity of anger, nor for anger-in, or anger-out expressions. Girls scored higher for attachmentanxiety to both mother and father, whereas boys scored higher for attachment avoidance tomother. Thus, girls were more likely than boys to be concerned about rejection and abandonmentby both mother and father figures, whereas boys were more likely than girls to be uncomfortable48with close relationships withtheir mothers and were more likely to strive tomaintain a sense ofindependence. However, it is importantto note that the effect sizes associated with these genderdifferences, as assessed by 2, were allweak, ranging from .000 to .021 in magnitude (Cohen,1988, suggests that ‘ti2 = .0099 refers tosmall effect, whereas 2 = .0588 refers to mediumeffect, andri2= .1379 refers to large effect).49Table 6Mean (Standard Deviation) ofPredictorand Outcome Scores by SexSexVariableBoys (n = 369) Girls (n 392) t-test dfii2Attachment anxiety to mother2.45 (0.79) 2.69 (0.87)4.00***759 0.02 1Scale range: I (low) to 7 (high)Attachment avoidance to mother334 (1.09) 3.11 (1.24)2.79**759 0.010Scale range: 1 (low) to 7 (high)Attachment anxiety to father 2.45 (0.90) 2.71 (0.93) 3.82*** 759 0.018Scale range: I (low) to 7 (high)Attachment avoidance to father 3.68 (1.26) 3.77 (1.38) 0.93 759 0.001Scale range: I (low) to 7 (high)Intensity of anger2.00 (0.58) 2.00 (0.57) -1.01 759 0.001Scale range: 1 (low) to 4 (high)Anger-in 2.20(0.55) 2.20 (0.54) 0.07 759 0.000Scale range: 1 (low) to 4 (high)Anger-out 2.04 (0.50) 2.00 (0.51) -1.09 759 0.002Scale range: I (low) to 4 (high)*<.05.**p<01***p<.001.50Tests ofHypothesesSeparate structural equating modeling(SEM) analyses were conducted, using the Mplus5.1 program (Muthén & Muthén, 2007),to examine the hypotheses of the present study for thefollowing six data sets: (a) mother figureattachment with the entire sample, (b) father figureattachment with the entire sample, (c) mother figureattachment with boys only, (d) motherfigure attachment with only girls, (e)father figure attachment with only boys, and (f) fatherfigure attachment with only girls. SEM allows usto determine the extent to which the theoreticalmodel hypothesized is supported by sample data. Thehypothesized model of the relationshipbetween attachment and anger dimensions wasestimated for each data set, specifying that allvariables were theoretically underlying continuousvariables. Given that the observed variableswere distributed reasonably multivaritate normal, maximumlikelihood (ML) estimation wasused in the SEM analyses, which assumes multivariate normal data.Model fit was evaluated with the comparative fit index (CFI), and root-mean-squareerrorof approximation (RMSEA), along withstandardized path coefficients for all of the proposedpaths. Results for the model-fit indices were presented in Table 6. Results forchi-square(x2)were also presented in that table, although the fitindex was not used in the actual evaluationbecause of its sensitivity to the size of the sample and correlations (Gerbing &Anderson, 1993).A reason for reporting the index in that table was that the formulas of mostif not all other indicesincludex2implying that it was a key ingredient (Kline, 2005). The CFI assesses the relativeimprovement in fit of the model compared with a baseline model, ranging from 0 to 1.0 (Bentler,1990). Larger CFI values indicate a better fit andvalues greater than .90 are considered a goodfit (Hu & Bentler, 1999). The RMSEA assesses the approximate fit of a model; howwell themodel would fit the population covariance matrix if it wereavailable (Browne & Cudeck, 1993).51RMSEA values less than .05 suggest agood fit and values ranging from .08 to .10 indicate areasonable or mediocre fit, and those greaterthan . 10 indicate a poor fit (Browne & Cudeck,1993; Hu & Bentler, 1999). Further,the mediator role of intensity of anger in the proposedmodel was examined by testing direct and indirecteffects in the structural model. Results of thefit indices are presented in Table 7and results of the direct and indirect effects are presented inTable 8. Path diagrams of the modelresults are also presented (see Figures 4 — 9).Analyses ofMother Figure Attachment with EntireSampleModel test. As seen in Table 7, the model was areasonable fit to the data for the motherfigure attachment with the entire sample, CFI = .99,RMSEA .07. Standardized pathcoefficients were significant for all hypothesized paths, except for the path fromattachmentavoidance to anger-out and the path from the attachment anxietyx attachment avoidanceinteraction5to intensity of anger. The coefficient for the path from attachment anxiety to anger-out was not significant, as predicted. As hypothesized, adolescents’ attachment anxiety andavoidance toward mother figures were positively related to the adolescents’ high level of angerfeelings (i.e., intensity of anger) which, in turn, was associated with increases in anger-in andanger-out expressions, with direct effects of attachment anxiety and attachment avoidance onanger-in but no direct effects of attachment anxiety and avoidance onanger-out. Therelationships between both attachment anxiety and attachment avoidance and intensity of angerwere significant, and in turn the relationships between intensity of anger and both anger-in andanger-out expressions were significant. The significance of these two segments of the paths fromattachment anxiety and attachment avoidance to anger-in and anger-out expressions indicatesThe continuous predictor variables were first centered by saving their standardized scores as newvariables, andthen product terms (i.e., interaction terms) were created between the predictor variables (i.e., attachment anxietyandattachment avoidance).52mediation. Accordingly, the mediationalstructure was further examined with the significancetests for indirect effects.Tests ofmediation. The significancetest statistic was created, dividing the indirect effectby its standard error and the resulting ratiowas then compared to the standard normaldistribution to test its significance (i.e., z= a*b/standarderror ofa*b;a represents a direct effectbetween an independent and a mediator, brepresents a direct effect between the mediator and adependent variable) (MacKinnon, Lockwood,Hoffman, West, & Sheets, 2002). First, the twosegments of the indirect paths from attachment anxiety toanger-in expression through intensityof anger were examined. The indirect effect of attachmentanxiety on anger-in through intensityof anger was statistically significant(= .08, z=5.69,p < .001), providing evidence ofmediation. Moreover, to examine whether the mediation waspartial orfull, the direct effectbetween attachment anxiety (i.e., the independent variable)and anger-in expression (i.e., thedependent variable) was evaluated. If the direct effect is significant, then themediation is partial.If the direct effect is not significant, then the mediation is full. Asseen in Tables 7 and 8, thedirect effect between attachment anxiety andanger-in was significant, 3 .lO,p<.01, indicatingpartial mediation.Next, the mediational relationship between attachment anxiety and anger-out throughintensity of anger was examined in the same manner as described for the relationship betweenattachment anxiety and anger-in. The indirect effect of attachment anxiety on anger-out throughintensity of anger was statistically significant, 3 = .15,z=6.35,p < .001, providing evidence ofmediation. To examine whether the mediation was partial orfull, a direct path betweenattachment anxiety (i.e., the independent variable) and anger-out expression (i.e., the dependent53variable) was evaluated. As seen in Tables 7 and 8, there was no significance found on the directpath, 3 = -.01, ns, indicating full mediation.Equivalently, the indirect effect of attachment avoidance on anger-in through intensity ofanger was statistically significant, f3 = .08, z = 5.33, p < .00 1, supporting mediation. Also, therewas the statistically significant direct path between attachment avoidance and anger-in, j3 = .20,p<.001, suggesting partial mediation (see Tables 7 and 8).When examining the mediational relationship between attachment avoidance and anger-out through intensity of anger, there was a significant indirect effect of attachment avoidance onanger-out through intensity of anger, f3 = .14, z = 5.84, p < .001, but no significant direct pathbetween attachment avoidance and anger-out,1= .03, ns, suggesting full mediation (see Tables 7and 8).In sum, these results showed the important role of intensity of anger as a mediatorbetween maternal attachment and anger expression among adolescents. Adolescents reportinghigh levels of attachment anxiety experienced a higher level of anger feelings (i.e., intensity ofanger) and that intensity of anger, in turn, increased their anger-in expression. Adolescentsreporting greater levels of attachment avoidance were also more likely to experience a greaterlevel of anger feelings, and the experience of anger intensity, in turn, enhanced anger-inexpression. A similar pattern was found on the paths from attachment anxiety and attachmentavoidance to anger-out. However, the latter pattern of results — statistically significant indirecteffects but not direct effects (i.e., full mediation) — represents a stronger demonstration for themediator effect, assuming correct directionality specification. Hence, intensity of anger hasshown its critical role as a mediator on the path from insecure maternal attachment (i.e., anxietyand avoidance) to both anger-in and anger-out expressions, but the magnitude of the anger-intensity role as a mediator was strongeron the path from insecure attachment to anger-outexpression than that to anger-in.5455Table 7Fit Indices and Standardized Path Coefficientsfor Models1. Mother 2. Father 3. Mother 4. Mother 5. Father 6. Fatherfigure figure figure figure figure figureattachment: attachment: attachment: attachment: attachment: attachment:Entire Entire Boys Girls Boys GirlsFit Indicesx29.08 13.98 13.00 0.94 10.66 4.84df 2 2 2 2 22CFI .99 .98 .97 1.00 .97.99RMSEA .07 .08 .08 .00 .08.06PathsAnxiety --> Anger intensity.22*** .16*** .17*** .28*** .16***Anxiety --> Anger-in .10.15***.03.18***.04Anxiety --> Anger-out -.01 .01 .01 -.02 .02.00Avoidance-->Angerintensity.20*** .21*** .24*** .19*** .22***Avoidance -->Anger-in.20*** .17*** .23*** .17*** .20***Avoidance --> Anger-out .03 -.05 -.01 .07 -.02-.08Anxiety x Avoidance--> Anger intensity -.04 -.08 .00 -.08-.02Anger intensity --> Anger-in37*** 39*** 33***.42***34***Anger intensity-->Anger-out.67*** .69*** .65*** .68***.65*** .72****p< .05.**p< .01.***p< .00156Table 8Direc4 Indirect, and Total Effects ofStudyVariables on Endogenous Variables1 .Mother figure attachment: 2. Father figure attachment: 3. Motherfigure attachment:Entire (N 775) Entire (N = 762) Boys (n = 379)Anger Anger- Anger- Anger Anger- Anger- Anger Anger- Anger-Variable intensity in outintensity in out intensity in outAnxietyDirecteffect.22***.**-.01.16***.11 .01.17*** .15***.01Indirect effect-- .08*** .15*** -- .06***.11*** -- .06**.11***Total effect.22***.18***.14***.16***.17***.12***.17***.21***.12**AvoidanceDirect effect.20*** .20***.03.21*** .17***-.05.15*** .23***-.01Indirect effect-- .08*** .14*** -- .08*** .14*** -- .05**Total effect.20*** .28*** .17*** .21*** .25*** .09** 15*** .28***.09Anger intensityDirect effect-- •37***.67***-- 39***.69***-- 33***.65***Indirect effect -- -- -- -- -- -- -- -- --Total effect-- 37***.67***-- 39***.69***-- 33***.65***Note. All effects in this table were standardized values.*p< .05.**p< .01.***p< .00157Table 8 (Continued)4. Mother figure 5. Father figure 6. Father figureattachment:attachment: Girls (n = 396) attachment: Boys (n = 369)Girls (n = 393)Anger Anger- Anger- Anger Anger- Anger- Anger Anger- Anger-Variable intensity in outintensity in out intensity in outAnxietyDirect effect.28***.03 -.02.16*** .18***.02.17***.04 .01Indirect effect-- .12*** .19*** -- .05** -- .08*** .12***Total effect.28*** .15** 17*** .16*** .23*** .12* .17*** .12* .13***AvoidanceDirect effect.24*** .17***.07.19*** .20***-.02.22*** 15***-.08Indirect effect—— .16*** -- .07*** .13*** -— .16***Total effect.24*** .27*** .23*** .19*** .27*** .11* .22*** .25***.08Anger intensityDirect effect-- .42*** .68***-- 34***65***-- 44***.72***Indirect effect -- -- ---- -- -- -- -- --Total effect-- .42*** .68***34***65***-- 44***.72***Note. All effects in this table were standardized values.*p< .05.**p< .01.***p< .00158Figure 4. An attachment-anger model of mother figure with entire sample:CFI = .99, RMSEA = .07.Figure 5. An attachment-anger model of father figure with entire sample:CFI = .98, RMSEA = .08.37***59Figure 6. An attachment-anger model of mother figure with boys only:CFI = .97, RMSEA = .08.Figure 7. An attachment-anger model of mother figure with girls only:CFI = 1.00, RMSEA = .00.ns -- -- ----Figure 8. An attachment-anger model of father figure with boys only:CFI = .97, RMSEA = .08.Figure 9. An attachment-anger model of father figure with girls only:CFI = .99, RMSEA = .06.60ns - - - - - - - - -.144***61Analyses ofFather FigureAttachment with Entire SampleModel test. A pattern similar to that shownfor mother figure attachment with entiresample was found for father figure.The model was a reasonable fit to thedata of father figureattachment with entire sample, CFI = .98,RMSEA = .08. Standardized path coefficientsweresignificant for all hypothesized paths,except for the path from attachment avoidance toanger-outand the path from attachmentanxiety x attachment avoidance interaction tointensity of anger.There was no significance of thepath from attachment anxiety to anger-out, as expected.Consistent with the results for motherfigure, adolescents’ insecure attachment (i.e., attachmentanxiety and avoidance) toward father was positivelyrelated to the adolescents’ greater level ofanger feelings (i.e., intensity of anger)which, in turn, was associated with increases in anger-inand anger-out expressions, with adirect effect of attachment anxiety on anger-in but no directeffects of attachment anxiety and avoidance onanger-out. The relationships between bothattachment anxiety and attachment avoidance and intensityof anger were significant, and in turnthe relationships between intensity ofanger and both anger-in and anger-out expressions weresignificant, suggesting mediation effects. Hence, themediational structure was further examinedwith the significance tests for indirect effects.Tests ofmediation. In terms of the two segmentsof the indirect paths from attachmentanxiety to anger-in expression through intensity ofanger, the indirect effect of attachmentanxiety on anger-in through intensity of anger wasstatistically significant(1= .06, z = 4.28,p <.001), providing evidence of mediation. The direct effectbetween attachment anxiety and angerin was significant, = .11,p< .001, indicating partial mediation.With regard to the mediational relationship betweenattachment anxiety and anger-outthrough intensity of anger, the indirect effect ofattachment anxiety on anger-out through62intensity of anger was statistically significant,13= .11, z = 4.50, p < .001, providing evidence ofmediation. The direct path from attachment anxiety toanger-out was not significant,13= .01, ns,indicating the full mediation.Regarding the indirect effect of attachment avoidance onanger-in through intensity ofanger, the mediated effect was statistically significant,13= .08, z = 5.38, p < .00 1, supportingmediation. Also, there was the statisticallysignificant direct path between attachment avoidanceand anger-in,13= .17,p< .001, suggesting partial mediation (see Tables 7 and 8).When examining the mediational relationship between attachmentavoidance and anger-out through intensity of anger, there was a significant indirecteffect of attachment avoidance onanger-out through intensity of anger, 13 = .14, z = 5.76, p < .001, but nosignificant direct pathbetween attachment avoidance and anger-out,1= -.05, ns, suggesting full mediation (see Tables7and8).In summary, these results were consistent withthe results for mother figure, supportingthe important role of intensity of anger as a mediatorbetween adolescents’ attachment to fatherfigure and anger expression. Specifically, intensity of anger has shown itscritical role as amediator on the path from adolescents’ paternalinsecure attachment (i.e., both anxiety andavoidance) to both anger-in and anger-out expressions.However, the magnitude of the angerintensity role as a mediator was stronger on the path from insecureattachment to anger-outexpression than that to anger-in.Analyses ofMother Figure Attachment by SexModel test. Separate SEM analyses, to investigate sex differences,were conducted forboys and girls in relation to their attachmentto mother figure. Results for boys were consistentwith those for the entire sample data described earlier. Themodel was a reasonable fit to the data63of mother figureattachment for boys, CFI = .97,RMSEA = .08. Consistent with theresults forthe entire sample data,boys’ insecure attachment (i.e.,anxiety and avoidance) toward motherfigure was positivelyrelated to the boys’ greater level ofanger feelings (i.e., intensity of anger)which, in turn, was associatedwith increases in anger-in and anger-outexpressions, suggestingmediation effects (see path coefficientsin Table 7). There was a direct effectof attachmentanxiety on anger-in but nodirect effects of attachment anxiety andavoidance on anger-out.Overall, results for girls were similarto those for boys and the entiresample although the modelfit to the data of girls was somewhatbetter than that of boys, CFI = 1.00, RMSEA = .00.Therewas, however, one noteworthydifference in results for girls, that is, that a significantdirecteffect of attachment anxiety on anger-inwas not found (Table 7).Tests ofmediation. Results forfurther examination with the significance tests forindirect effects for boys were consistentwith results for entire sample. When examining thetwosegments of the indirect paths fromattachment anxiety to anger-in expression through intensityof anger for boys, the indirecteffect of attachment anxiety on anger-in through intensityof angerwas statistically significant(= .06, z = 3.02,p< .0 1), providing evidence of mediation. Thedirect effect between attachment anxietyand anger-in was significant, (3=.l5,p<.00 1,indicating partial mediation.With regards to the mediational relationshipbetween attachment anxiety and anger-outthrough intensity of anger for boys, theindirect effect of attachment anxiety on anger-outthrough intensity of anger was statisticallysignificant,I= .11, z = 3.25, p < .001, providingevidence of mediation. The directpath from attachment anxiety to anger-out was not significant,3= .01, ns, indicating full mediation.64In terms of the indirect effect of attachmentavoidance on anger-in through intensity ofanger for boys, the mediated effectwas statistically significant, f3 = .05,z = 2.76,p< .01,supporting mediation. Also, there wasthe statistically significant direct path betweenattachmentavoidance and anger-in, 3=.23,p < .001, suggesting partial mediation (see Tables7 and 8).Finally, when examining themediational relationship between attachment avoidance andanger-out through intensity of anger for boys,there was a significant indirect effect ofattachment avoidance on anger-out throughintensity of anger,13= .10, z = 2.92,p<.00 1, but nosignificant direct path between attachmentavoidance and anger-out,1= -.01, ns, suggesting fullmediation (see Tables 7 and 8).These mediated effects for boys were consistent withresults for girls, except that themediation between attachment anxiety and anger-inwas full for girls instead of partial,representing the stronger demonstration for the mediatoreffect as compared to that for boys.This stronger mediator effect for girls correspondswith the non-significance of the direct pathbetween attachment anxiety and anger-in presented earlierfor girls. The statistical values forcoefficients (i.e., (3s) and significance tests (i.e., zs) forgirls were as follows: (a) (3 = .12, z = 5.11,p < .001 for the indirecteffect and (3 = .03, ns for the direct effect between attachment anxietyand anger-in (i.e., full mediation); (b)1=.19, z=S.84,p < .001 for the indirect effect and(3 = -.02, ns for the direct effect between attachment anxietyand anger-out (i.e., full mediation); (c)13= .10, z= 4.55,p< .001 for the indirect effect and (3=.l7,p < .001 for the direct effect betweenattachment avoidance and anger-in (i.e., partial mediation);and (d) (3 = .16, z=5.O3,p < .001 forthe indirect effect and13= .07, ns for the direct effect between attachment avoidance andangerout (i.e., full mediation).65In summary, the mediatedeffects were overall similar for boys and girls,bothdemonstrating a critical role of angerintensity as a mediator. Consistent with the resultsforentire sample, intensity of angerhas performed a critical mediator rolebetween adolescents’maternal insecure attachment (i.e.,both anxiety and avoidance) and both anger-inand anger-outexpressions, though the magnitude of theanger-intensity role as a mediator, again, wasstrongeron the path from insecureattachment to anger-out expression than that to anger-in.However,there was a notable difference for boysand girls, in that intensity of anger demonstrated astronger mediator effect on the pathfrom attachment anxiety to anger-in for girls than for boys.Analyses ofFather Figure Attachment by SexModel test. Separate SEM analyses were conductedfor boys and girls to examine sexdifferences in the hypothesized model in relationto their attachment to father figure. Results forboys were, again, consistent with those for theentire sample data described earlier. The modelwas a reasonable fit to the data of father figure attachment for boys,CFI = .97, RMSEA = .08.Consistent with the results for the entire sample data,boys’ insecure attachment (i.e., anxiousand avoidant) toward father figure waspositively related to the boys’ higher level of angerfeelings (i.e., intensity of anger) which, in turn, wasassociated with increases in anger-in andanger-out expressions, suggesting mediation effects (see pathcoefficients in Table 7). There wasa direct effect of attachment anxietyon anger-in but, again, no direct effects of attachmentanxiety and avoidance on anger-out. This pattern wasconsistent with the results described forentire sample and boys with maternal attachment.Overall, results for girls were similar to those for boysand the entire sample although themodel fit to the data of girls was, again, somewhatbetter than that of boys, CFI = .99, RMSEA.06. Notable differences in resultsfor girls were that: (a) there was no significant direct effect of66attachment anxiety on anger-in, consistent with the results in mother figure attachment for girlspresented earlier, and (b) a negative and significant direct effect of the interaction, attachmentanxiety x attachment avoidance, was found (see Table 7), suggesting a buffering effect of eitherattachment anxiety or attachment avoidance on the relationship between either of thoseattachment dimensions and intensity of anger. As a follow-up to the significant interaction, asimple main-effects-like analysis was conducted, wherein the models were fit for adolescentswho reported low versus high levels of attachment anxiety or avoidance, base on the median split(Steiger, 2003). Results of this follow-up analysis showed that, for adolescents who reportedhigh levels of attachment avoidance6,the effect of attachment anxiety was no longer significantfor increases in the intensity of anger. The comparisons of the main effect of attachment anxietyfor low and high attachment avoidance are presented Table 9.6In the present study, the low and high attachment groups were created based on attachment avoidance.Alternatively, in lieu of the attachment avoidance, attachment anxiety can be used to create low and high groups ofattachment anxiety.67Table 9Fit Indices and Standardized Path Coefficientsfor Models ofFather Figure Attachmentfor GirlsLow attachment High attachmentavoidance to father avoidance to father(n 188) (n = 205)Fit Indicesx20.79 0.41CFI 1.00 1.00RMSEA .00 .00PathsAnxiety --> Anger intensity.31***.05Anxiety --> Anger-in .11(p= .07) -.03Avoidance --> Anger intensity .06.15*Avoidance -->Anger-in .08 .12(p= .07)Avoidance --> Anger-out .04 -.06Anger intensity --> Anger-in.52*** .36***Anger intensity --> Anger-out.70****p< .05.**p< .01.***p< .00168Tests ofmediation.Results of analyses for indirecteffects for boys were consistent withresults for entire sample.When examining the two segmentsof the indirect paths fromattachment anxiety to anger-inexpression through intensity of angerfor boys, the indirect effectof attachment anxiety onanger-in through intensity of angerwas statistically significant ( .05,z=2.93,p < .001), confirmingevidence of mediation. Thedirect effect between attachmentanxiety and anger-in wassignificant,13= .18,p< .001, indicating partial mediation.When testing the mediationalrelationship between attachment anxietyand anger-outthrough intensity of anger forboys, the indirect effect of attachment anxietyon anger-outthrough intensity of anger was statisticallysignificant,13= .10, z = 3.13,p< .01, providingevidence of mediation. The directpath from attachment anxiety to anger-out was not significant,13= .02, ns, suggesting fullmediation.In terms of the indirect effect ofattachment avoidance on anger-in throughintensity ofanger for boys, the mediated effectwas statistically significant,13= .07, z=3.78,p < .001,supporting mediation. Also, therewas the statistically significant direct path between attachmentavoidance and anger-in,13.20,p< .001, indicating partial mediation (see Tables 7 and8).Finally, when examining the mediationalrelationship between attachment avoidance andanger-out through intensity of anger for boys,there was a significant indirect effect ofattachment avoidance on anger-out throughintensity of anger,13= .13, z 3.68,p<.001, but nosignificant direct path between attachmentavoidance and anger-out,1 =- .02, ns, indicating fullmediation (see Tables 7 and 8).These mediated effects for boys were consistentwith results for girls, except that themediation between attachment anxietyand anger-in was, again, was found to be full for girlsinstead of partial, representingthe stronger demonstration for the mediator effect ascompared to69that for boys. The statistics values for coefficients (i.e.,13s)and significance tests (i.e., zs) forgirls were as follows: (a) = .08, z=3.40,p < .001 for the indirect effect and13= .04, ns for thedirect effect between attachment anxiety and anger-in(i.e., full mediation); (b)13= .12, z = 3.56,p < .001 for the indirect effect and13= .01, ns for the direct effect between attachmentanxietyand anger-out (i.e., full mediation); (c)13= .10, z 4.28,p < .001 for the indirect effectand f3 =.15, p < .001 for the direct effect between attachment avoidanceand anger-in (i.e., partialmediation); and (d)13= .16, z=4.5l,p < .001 for the indirecteffect and13-.08, ns for thedirect effect between attachment avoidance and anger-out(i.e., full mediation).To sum up, overall the mediated effects were similarfor boys and girls, with intensity ofanger playing an important mediator in both cases. The resultsfor the entire sample wereconsistent in that intensity of anger was a critical mediatorfor the path between adolescents’paternal insecure attachment (i.e., both anxiety andavoidance) and both anger-in and anger-outexpressions. The level of the anger-intensity role as a mediator,again, was stronger on the pathfrom insecure attachment to anger-out expression thanthat to anger-in. However, there was anoteworthy difference for boys and girls. Intensity of angerdemonstrated a greater mediatoreffect on the path from attachment anxiety to anger-infor girls than for boys. This result wasconsistent with the result for maternal attachment.Relative Impact ofAttachment Dimensions and FiguresGiven that the theoretically-derived model, based onBowiby’s hypothesis, wassuccessfully supported by analyses in the present studyin a relatively large sample ofadolescents, the following analyses were conducted in orderto investigate the relative impact ofthe attachment dimensions and figures on anger (i.e., Whichattachment dimension has moreimpact on increases in anger, attachment anxiety or attachmentavoidance? Which attachment70figure has more impact on increasesin anger?). To achieve the objective, regression analyseswere conducted and values of Pratt index werecalculated.Prior to the primary regression analyses,multicollinearity among predictors (i.e.,attachment anxiety to mother, attachment avoidanceto mother, attachment anxiety to father,attachment avoidance to father, and intensity of anger) was diagnosed with toleranceand thevariance inflation factor (VIF) statistic indices.Following Kline (2005) and Miles and Shevlin’s(2001) recommendations, the cut-off point of .10 for tolerance and 4 for VIFwere used. That is,a value of tolerance<. 10 and a value of VIF>4 indicate a problem of multicollinearity. As seenin Table 10, none of the predictor variables was at risk of multicollinearity. Results ofbivariatecorrelations among relevant variables are also presented in Table 11.In addition, to examine whether sex of adolescents moderates the relationship betweenattachment dimensions and anger, a hierarchical regression test was conducted, regressing eachof the anger variables (i.e., intensity of anger, anger-in, anger-out) on the centered attachmentvariables (i.e., attachment anxiety to mother, attachment avoidance to mother, attachment anxietyto father, and attachment avoidance to father), and sex of adolescents in the first step and addingall the two-way product terms of each of the attachment variables by sex in step two. The resultsof the regression analyses are presented in Tables 12 - 14. First, for intensity of anger, the firststep accounted for 12% of the variance in intensity of anger. However, the change in R2 = .01 forthe second step was not significant (p = .15), indicating that sex did not moderate the relationshipbetween the attachment variables and intensity of anger (Table 12). The same pattern wasobserved for anger-in (see Table 13) and anger-out (see Table 14) variables. Given no significantsex differences, the following primary analyses for this section were performed only on theentire sample. Although the change in R2 for the second step was not significant, there was a71significant interactionbetween attachment avoidance to mother and sex ofadolescents in relationto anger-out expression, = -.31,p< .05 (see Table 14). Further investigation of the significantinteraction is beyond the scope of thepresent study. However, it should be noted thatthissignificant interaction suggests amoderation effect of sex on the relationship between attachmentavoidance to mother and anger-out expression.72Table 10Results ofMulticollinearity Diagnosisfor Predictor VariablesVariable Tolerance VIEEntire sample (N 761)Attachment anxiety to mother .53 1.90Attachment avoidance to mother.75 1.33Attachment anxiety to father.55 1.83Attachment avoidance to father.74 1.35Intensity of Anger.89 1.13Boys (n = 369)Attachment anxiety to mother .41 2.46Attachment avoidance to mother.63 1.53Attachment anxiety to father.42 2.38Attachment avoidance to father.60 1.66Intensity of Anger .93 1.08Girls (n = 392)Attachment anxiety to mother .61 1.63Attachment avoidance to mother.80 1.26Attachment anxiety to father.64 1.56Attachment avoidance to father.82 1.23Intensity of Ancier .831.2173Table 11Intercorrelations Among Study Variables1 2 3 4 5VariableEntire sample (N = 761)1. Attachment anxiety to mother2. Attachment avoidance to mother3. Attachment anxiety to father4. Attachment avoidance to father5. Intensity of anger6. Anger-in7. Anger-outBoys (n = 369)1. Attachment anxiety to mother2. Attachment avoidance to mother3. Attachment anxiety to father4. Attachment avoidance to father5. Intensity of anger6. Anger-in7. Anger-outGirls (n 392)1. Attachment anxiety to mother2. Attachment avoidance to mother3. Attachment anxiety to father4. Attachment avoidance to father5. Intensity of anger6. Anger-in7. Anger-out*p<05**p<.01.1.00I3**67**19**24**21**1.0014**75**21**I8**25**.13*1.0014**59**I6**30**19**1.0013**1 0048** 12**1 0023** 18** 24**1 0030** 19** 28**45**1 00.19.12** .67** .30** 1.0060**09 1 0017** 17** 21**1 0031** 25** 28** 41**1 00.09.13* .12* .65** .23**1.001.0019**1 0040** 14**1 0029** 19** 26**1 0030** 14** 27**49**1 00.26** .13* .11* .69**37**1.0074Table 12Summaiy ofTestfor ModerationEffects ofSex on the Prediction ofIntensityofAnger (N = 761)Variable BSEBStep IAttachment anxiety to mother 0.110.03Attachment avoidance to mother0.08 002Attachment anxiety to father0.01 0.03 .02Attachment avoidance to father0.08 0.02Sex of adolescents0.07 0.04 .06Step 2Attachment anxiety to mother 0.26 0.08Attachment avoidance to mother0.18 0.07Attachment anxiety to father-0.09 0.08 -.16Attachment avoidance to father0.07 0.07 .11Sex of adolescents0.07 0.04 .06(Attachment anxiety to mother) x (Sex of adolescents) -0.11 0.06 -.29(Attachment avoidance to mother) x (Sex ofadolescents) -0.07 0.05 -.19(Attachment anxiety to father) x (Sex of adolescents)0.08 0.06 .22(Attachment avoidance to father) x(Sex of adolescents) 0.01 0.05.04Note. R2= .12, F(5, 755) = 19.73,p< .001 for Step 1; AR2= .01, AF(4, 751) = 1.71,p= .15 for Step2.*p< .05.**p< .01.***p< .00175Table 13Summaiy of Testfor Moderation Effects ofSexon the Prediction ofAnger-In (N = 76])Variable B SE BStep 1Attachment anxiety to mother 0.06 0.03 .11**Attachment avoidance to mother0.11 0.02Attachment anxiety to father0.04 0.03 .08Attachment avoidance to father 0.080.02Sex of adolescents0.01 0.04 .01Step 2Attachment anxiety to mother 0.10 0.08 .18Attachment avoidance to mother 0.100.06 .18Attachment anxiety to father -0.100.08 -.18Attachment avoidance to father 0.100.06 .18Sex of adolescents 0.01 0.04.01(Attachment anxiety to mother) x (Sex of adolescents) -0.03 0.05 -.09(Attachment avoidance to mother) x (Sex of adolescents) 0.010.04 .03(Attachment anxiety to father) x (Sex of adolescents)0.10 0.05 .28(Attachment avoidance to father) x (Sex of adolescents) -0.010.04 -.13Note. R2= .14, F(5, 755) = 25.04,p< .001 for Step 1; AR2= .01, tiF(4, 751) = 1.15, p = .33 for Step2.*p< .05.**p< .01.***p< .00176Table 14Summary of Testfor ModerationEffects ofSex on the Prediction ofAnger-Out(N 761)VariableB SE BStep 1Attachment anxiety to mother 0.070.03Attachment avoidance to mother0.08 0.02Attachment anxiety to father0.01 0.02 02Attachment avoidance to father0.01 0.02 .01Sex of adolescents0.05 0.04 .04Step 2Attachment anxiety to mother 0.15 0.07.30*Attachment avoidance to mother0.23 0.06Attachment anxiety to father-0.07 0.07 -.13Attachment avoidance to father-0.06 0.06 -.11Sex of adolescents0.05 0.04 .05(Attachment anxiety to mother) x(Sex of adolescents) -0.07 0.05 -.19(Attachment avoidance to mother)x (Sex of adolescents) -0.11 0.04(Attachment anxiety to father) x (Sex of adolescents)0.06 0.05 .17(Attachment avoidance to father) x(Sex of adolescents) 0.05 0.04 .15Note. R2= .06, F(5, 755) = 9.02,p< .001 for Step 1; zR2= .01, tF(4, 751) = 1.84,p= .12 for Step2.*p< .05.**p< .01.***p< .00177To assess therelative impact of attachmentdimensions and figures on anger,relative-importance values for predictorvariables were calculated,using the Pratt index (Thomas,Hughes, & Zumbo, 1998;Thomas & Zumbo, 1996)in regression analyses. The Prattindexinforms us of the contribution ofevery predictor variable studiedto the overall R2, ordering thepredictor variables in termsof the fraction of the R2which is attributed to each predictorvariablein a model. This value is theproduct of the bivariate correlation andthe beta weight divided bythe R2 (i.e.,r*/R2). Results are presentedin Tables 15 — 17.As seen in Table 15,attachment anxiety to mother (i.e., Prattindex value = .40) was mostinfluential for contributingto increases in the intensity of anger,followed by attachmentavoidance to mother (Pratt = .30),attachment avoidance to father (Pratt.27), respectively.When looking at the relativecontribution based on attachment figures,mother figure contributedmore to increases in the intensityof anger than father figure,Pratt = .70 for mother figure; Pratt= .29 for father figures. When weassessed the contribution basedon attachment dimensions,attachment avoidance was more influentialthan attachment anxiety contributing to highlevels ofanger feelings, Pratt = .57 for avoidance;Pratt .42 for anxiety.For anger-in (see Table 16),attachment avoidance to mother (Pratt = .45)was the mostdominant variable, followed by attachmentavoidance to father (Pratt = .28), attachmentanxietyto mother (Pratt = .16), andattachment anxiety to father (Pratt = .10), respectively.Mother figurewas more influential than fatherfigure (Pratt = .61 for mother; Pratt= .38 for father), andattachment avoidance contributedmore than attachment anxiety (Pratt = .73for attachmentavoidance; Pratt = .26 for attachmentanxiety). When the variable of anger intensitywas includedin the model, it was found that theintensity of anger was the strongestvariable, but the order ofthe relative contributionfor the rest of the variables (i.e., attachmentanxiety to mother,78attachment avoidance tomother, attachment anxiety tofather, and attachment avoidance tofather) was consistent withthose when the intensityof anger was excluded from the model. Thereduced levels of the Prattindex values for the attachment variablesafter adding the variable ofthe anger intensity might becorrespondent with the results of the meditationalrole of angerintensity found from the SEM analysespresented in the previous sections.In terms of anger-out (seeTable 17), attachment avoidance to mother(Pratt = .59) wasmost contributable, followedby attachment anxiety (Pratt = .37), attachmentanxiety to father(Pratt = .03), attachment avoidanceto father (Pratt = .01). Mother was considerablya strongerfigure than father (Pratt = .96 formother; Pratt .04 for father), and attachment avoidancewasmore contributable than attachmentanxiety (Pratt = .60 for mother; Pratt = .40 for father).Whenthe variable of anger intensitywas added to the model, the intensity of angerwas, again, thestrongest, reducing the levels of thePratt index values of the attachment variables.In summary, mother figurecontributed more than father figure to increasing levels ofadolescents’ anger feelings and anger-in andanger-out expressions. Among adolescents,attachment avoidance was also consistentlymore influential for contributing toincreases inanger intensity and expressionacross all anger variables than attachmentanxiety.79Table 15Summary ofRegression Analysisfor Variables Predicting Intensity ofAnger (N = 761)VariableB SE BCorrelation PrattModelAttachment anxiety tomother 0.13Attachment avoidance tomother 0.07Attachment anxiety to father0.01Attachment avoidance to father0.06Note. R= .11, F(4, 756)23.91, p <.001, for the model.*p<.05.**p<.01.***p<.00100319***24 400.02.15***.23 .300.03 .02 .***.23 .2780Table 16Summaiy ofRegression Analysisfor Variables Predicting Anger-In (N = 761)Variable B SE B Correlation PrattModelAAttachment anxiety to mother 0.07 0.03.11*.21 .16Attachment avoidance to mother 0.10 0.02.21***.30 .45Attachment anxiety to father 0.04 0.02 .08 .19 .10Attachment avoidance to father 0.06 0.01 .27 .28Model BAttachment anxiety to mother 0.02 0.03 .04.21 .03Attachment avoidance to mother 0.07 0.02.16***.30 .18Attachment anxiety to father 0.04 0.02 .07 .19 .05Attachment avoidance to father 0.04 0.01.10**.27 .10Intensity of Anger 0.34 0.03.36***.45 .63Note. Model A: R2= .14, F(4, 756)= 31.33,p<.001; Model B R2= .26, F(5, 755)= 53.31,p<.001*p< .05.**p< .01.***p< .00181Table 17Summaiy ofRegression Analysisfor Variables Predicting Anger-Out (N= 761)Variable B SEBf3 Correlation PrattModel AAttachment anxiety tomother 0.08 0.03.13**.16 .37Attachment avoidance tomother 0.07 0.02j7***.19 .59Attachment anxiety to father0.01 0.03 .01 .12 .03Attachmentavoidancetofather0.00 0.01 .01 .11.01Model BAttachment anxiety to mother 0.00 0.02 .00 .16.00Attachment avoidance to mother*.19 .03Attachment anxiety to father0.00 0.02 .00 .12.00Attachment avoidance to father -**.11Intensity of Anger 0.600.02.67***.67 .97Note. Model A: R2= .05, F(4, 756) =10.89,p< .001; Model B: R2= .46, F(5, 755) = 129.07,p< .001*p< .05.**p< .01.***p< .001The small negative value of the Pratt indexessentially indicates zero (Thomas, Hughes, & Zumbo, 1998).82DiscussionSummary and Discussion ofFindingsThis study examined therelationship between attachment and anger amongadolescentsas predicted from an attachmentperspective, addressing a long-standing,but as yet untestedprediction that attachment to caregivershas implications for how individuals experience andexpress anger. Specifically, thepurpose of this study was to investigate: (a) the natureof therelationship between attachmentdimensions and anger experience and expression, (b) whetherthere are sex differences in theattachment-anger link; and (c) whether the relationshipbetweenattachment and anger differs between attachment figures,mother and father. This workaddressed these research questions byproviding evidence from a large sample of adolescents,with a theoretically and methodologically refinedtest of the hypothesis. Specifically, the presentstudy tested a model of anger whichwas carefully structured based on an attachment perspectiveas originally hypothesized by John Bowlby (1973).This study utilized a micro-approach to investigate acritical component of social-emotional development, anger, in relation to attachment. A macro-approachwhich is oftenconsidered “ecological”, as opposed to a micro-approach,has been increasingly popular ininvestigating children and adolescents’ social-emotionalwell-being. Among researchers macroapproaches are often used to investigate protective andrisk factors, examining how individual,family, school and/or community factorsare likely to contribute to outcome variables. Althoughmacro approaches are useful for drawing a broader pictureof protective and risk factorssurrounding children and adolescents, exclusive relianceon macro approaches may lead tooverlooking a crucial local factor or situation. Both micro-and macro-approaches are valuableand necessary and should be embraced within a dynamicand nested ecological approach.83Confirming Bowlby’s (1973)original hypothesis, the fmdings fromthe present studydemonstrated that adolescents’attachment to their caregivers predictsboth anger experience andexpression. Unlike previousresearch (Kobak et al., 1993; Kobak& Sceery, 1988), in thepresentstudy, anger experience/feelingand anger expression were differentiatedas critical dimensionsof anger. This careful approachto investigating how differentdimensions of anger are predictedby different dimensions ofattachment resulted in discovering acritical role of experience/feelings of anger (i.e., intensity of anger)as a mediator on the path from attachmentto angerexpression.Specifically, the present studyfound that adolescents having high levels of attachmentanxiety and attachment avoidancereported experiencing a greater amount of angerfeelings, andthe increases in the intensity oftheir felt anger, in turn, elevated the level of both anger-inandanger-out expressions. Thisremarkable finding of mediation might help to explainwhy previousstudies (Calamari & Pini, 2003; Mikulincer, 1998)have shown mixed results for the attachment-anger relationship. Whereas resultsof Mikulincer’ s study indicated that ambivalently/anxiouslyattached college students were more likely toexpress anger inwardly than were studentswithsecure attachment, Calamari andPini found that adults with avoidant attachment weremorelikely to express anger inwardly than werethose with secure attachment. By solely relyingonregression analyses, these researchers failed tointegrate another important element of anger,experience/feelings of anger (i.e., intensityof anger), into their research model of attachment andanger. Contrary to the regression approach,structural equation modeling (SEM) allowed us toexamine relationships among multipleindependent and dependent constructs simultaneouslyandsequentially, in contrast to regressionmodels which perform analyses of only one layer oflinkages between independent and dependentvariables at a time.84Despite the overall finding of the mediationally-structured relationship between insecureattachment and anger, there was a notable difference found between the path from insecureattachment to anger-in versus anger-out expression as a function of anger intensity.Adolescentswho reported high levels of attachment anxiety as well as those who reported high levels ofattachment avoidance more likely to report expressing anger inwardly and this was partiallymediated by levels of intensity of their anger. Adolescents who reported greater levels of anxiousattachment as well as adolescents who reported greater levels of avoidant attachment alsoreported greater tendency to express their anger outwardly, and this relationship was fullymediated by their level of anger intensity. In other words, experiencing greater levels of anger isa necessary factor to direct anxious and avoidant adolescents toward a higher level of outwardly-expressed anger.Although it was hypothesized that there would be a direct effect of attachment avoidanceon anger-out expression given Kobak and Sceery’s (1988) finding of an association betweendismissing (or avoidant) attachment and hostility-type of anger rated by peers, this direct effectof attachment avoidance on anger-out expression was not found in the present study. Onepossible explanation for these discrepant findings is that Kobak and Sceery used a very generalmeasure of anger that tapped hostility more than anger per se and failed to consider differentdimensions of anger. Anger represents feelings, whereas hostility often refers to negativeattitudes and destructive and punitive behavior (Spielberger, 1999). Bowlby’s hypothesis(1969/1997, 1973) of an attachment-anger link concerned the experience/feelings of anger, nothostility, although these concepts overlap to an extent. Furthermore, Kobak and Sceery utilized apeer rating assessment of anger that may be less sensitive to inner process of anger. It might be85relatively easy for peers toobserve other’s outwardly-expressed anger,but it would be difficultfor them to identify someone’sinner feelings of anger.The present results, demonstratingthe attachment-anger link mediated by anger intensity,were consistently observed in bothattachment figures, mother and father. The relevantparent inprevious attachment studies has been solely the mother,and far less is known about the putativeinfluence of child-father attachment, withno known empirical studies to date examining therelationship between child-father attachment and anger.In the present study, the theoreticallyconstructed model was a reasonable fit tothe data for both mother and father figures, indicatedby model-fit indices. The mediationalstructure of the model was also supported for father figureattachment; that is, adolescents who were more insecurelyattached to father (i.e., reporting highlevels of attachment anxiety and high levelsof attachment avoidance) were more likely toexperience a higher intensity of anger feelings which, in turn, increasedthe level of expressingtheir anger in aggressive behavior. This was also truefor suppressed anger (inwardly-expressedanger).Overall, results were consistent across attachment figures (mother,father) when analyzedseparately. Specifically, for both mothers and fathers, the present resultsindicated partialmediation for the path from attachment anxiety and attachment avoidance toanger-in and fullmediation for the path from attachment anxiety and attachment avoidance to anger-out.Thisconsistent finding for mother and father figures underscores the importanceof both parents in theattachment process. Conventional emphasis has paid more attention to mother figurein theliterature in terms of their impact on their children’s well-being. Subsequentfindings indicatingthat attachment to both mother and father figures does have an impact onhow adolescents86experience and deal with their anger expands on the attachment literature by adding the criticalinfluence of father figure on adolescents’ anger process.With regard to sex differences in the attachment-anger relationship, a reasonable to goodmodel fit and a similar overall mediational structure was found in the present study for girls andboys with respect to both mother and father figures, consistent with the results observed in theentire sample. There was, however, one interesting difference between boys and girls. Whereasthe mediation between attachment anxiety and anger-in was partial for boys, consistent with theresults from the entire sample, the mediation for girls was full. This was true for both mother andfather figures. This full mediation for girls apparently resulted from non-significant direct effectof attachment anxiety on anger-in expression and suggests that a more intense angry feelings is anecessary factor to direct girls reporting high levels of attachment anxiety toward a propensity tosuppress their anger. In other words, without experiencing more intense angry feelings, there isno relationship between attachment anxiety and anger-in among girls. Simply having attachmentanxiety to their caregivers does not lead to a greater amount of anger suppression (anger-in) forgirls.In previous studies of adult populations, sex differences in assessed anger have not beenconsistently found; some have reported significant sex differences (Ben-Zur & Zeidner, 1988)whileothers have not (Bartz et al., 1996; Kopper, 1993; Newman et aL, 1999). As well, there isno empirical research providing convincing evidence of sex differences in anger expression, atleast in adult populations, although authors in the theoretical literature appear to contend sexdifferences in anger expression (i.e., women are more likely to mask or suppress their feelings ofanger than men; Chaplin, 2006; Fischer et al., 1993; Kopper & Epperson, 1996; Newman et al.,1999; Sharkin, 1993). One study (Cox et al., 2000) in a child and adolescent population,87however, has demonstrated significant sexdifferences in anger expression. In their study, withschool children from grades 5-9, Cox et al. found thatgirls reported higher levels of anger-inexpression than boys. However, in the present study, with an adolescentpopulation (i.e., an olderpopulation than the population examined in Cox et al. ‘s study), significant sex differenceswerenot found on anger-in as well as anger-out expression. This differentresult between the presentstudy and Cox et al. ‘s study may reflect a developmental difference that“boys lag behind girls”(Jacobs, Phelps, & Rohrs, 1989,p.64). Research (Jacobs et al., 1989; Chaplin, Cole, & ZahnWaxler, 2005; Brody & Hall, 2008) has suggested that boys learn to beless expressive ofemotions, including anger, as they develop, catching up with girls. Considering thisdevelopmental point, the male adolescents who participated in the present study might be morelikely to have learned being less expressive of anger feelings than the younger participants inCox et al.’s study. As a result, there was no longer sex difference in the present study. To date,there are no studies examining sex differences in the relationship betweenattachment and anger.Further research is needed to replicate this finding.Finally, with regard to sex differences in the present study, there was a significantinteraction between attachment anxiety and attachment avoidance with fathers on the intensity ofanger for girls not for boys or for the entire sample. This difference for girls might have beenderived from the results of the exploratory factor analyses of the attachment anxiety variable forgirls. That is, the variable of attachment anxiety was marginally within acceptable range ofunidimensionality for the data set of father figure attachment with girls. Accordingly, there isroom to speculate a possibly weak construct comparability of the attachment anxiety scalebetween girls and boys, with a clear need for further research in this area.88The present study utilized a dimension-basedattachment measure that was designed totap adolescents’ on-goingattachment to their caregivers. Given theoretical and empiricaldemonstrations of attachment stability throughoutlife, use of a self-report measure of currentattachment to parents was considered appropriate forinvestigating the role of child-parentattachment on a mechanism of anger. Among the small number ofprevious studies that haveinvestigated the attachment-anger relationship, themajority of them (Calamari & Pini, 2003;Meesters & Muris, 2002; Mikulincer, 1998; Murid et a!., 2004;Troisi & D’Argenio, 2004) haveassessed attachment in romantic relationships,tapping attachment to intimate partners inromantic relationships, not to parents. Utilizing a newly developed self-report measure ofadolescent attachment to parents, the present study not only provided direct evidence to supportBowiby’s hypothesis of dysfunctional anger as a predictable correlate of insecure attachment, butalso provided evidence that these relationships are evident among adolescents and are similar forboth boys and girls and with regard to attachment to both mothers and fathers.A dimensional approach to assessing attachment patterns was useful in the present studyfor several reasons. Conceptualizing attachment patterns in dimensional termswere morereasonable and practical, as Fraley and Wailer (1998) reported that there is no evidence tosupport a true attachment typology, and that the conceptual styles of attachment are regions in adimensional space. Use of a typological measure instead of a dimensional scale may lead us tolosing precision of examining a research objective in relation to attachment patterns. Finally, asdescribed earlier, a dimensional understanding of attachment patterns has been empiricallysupported by Brennan et al. (1998).As an extension to the SEM analyses to test the attachment-anger model developed in thepresent study, regression analyses were employed to identify relative contributions of attachment89dimensions and figures on anger. Due to theinsufficient sample size, the present study was notable to directly assess therelative impact of father and mother figures in relation toangerexperience and expression simultaneouslyin one model. Instead, an examination of the relativecontributions of attachment dimensions and figuresto anger was attempted by using the Prattindex in regression models. Resultsindicated that mother attachment contributed more to angerintensity and expression than fatherattachment, and that attachment avoidance was moreinfluential than attachment anxiety in predicting levelsof anger intensity and anger-in and anger-out expressions. The strongercontribution of the mother figure is consistent with Bowlby’snotion of monotropy which specified that a biological motherwho has the greatest biologicalinvestment in her child is likely to be most influential in thedevelopment of the child (Bowlby,1967/1997; Cassidy, 2008). Regression analyses examiningthe relative contribution ofattachment figures in the present study provide an initialidea of the relative impact of motherand father figures. Specifically, mother figure contributed more than father figure toincreasinglevels of adolescents’ anger feelings and anger-in and anger-outexpressions. These results,however, cannot tell us whether there is a statistically significant differencein the relativecontributions of mother versus father attachment. Further research with alarger sample size isneeded to examine the relative impact of attachmentfigures within the attachment-anger model.Implications ofFindingsA number of important implications emerged from the present study. First, the findingthat insecure attachment contributes to predicting greater feelingsof anger and unhealthyexpressions of anger serves as a strong reminder of the significant role ofcaregivers onadolescents’ well-being and current social-emotional functioning. Indeed,conventional wisdomtends to stress the importance of peer influences on individuals’well-being during the period of90adolescence. Peer approvalbecomes increasingly critical as one moves intolater childhood andbeyond (e.g., Harter, 1990).However, peer influence is not the only majorfactor affectingadolescent life. Researchers (e.g., Harter, 1990,1999) acknowledge that parental approvalcontinues to significantly impact onone’s well-being in her or his later life. For example,in theself-understanding literature, Harter (1999)found that correlations between parent approval andself-esteem did not decline, although teacherand peer approval became more predictive duringmiddle childhood than during early childhood andthis trend continued into adolescence. Theresults of the present study strengthenthe considerable role of caregivers on individuals’ growtheven during adolescence.Another point to be addressed with regard to the significant role of caregivers onadolescents’ well-being is a deleterious nature of anger as suggested byclinical and healthliterature (Biaggio & Godwin, 1987; Blumberg & Izard,1985; Chaplin, 2006; Farmer, 2002;Moreno et al., 1993; Pipher, 1994; Riley et al., 1989;Robbins & Tanck, 1997; Seidlitz et al.,2000). Researchers and practitioners in the clinical andhealth fields have provided extensiveevidence for the predictive association of anger, especially anger expression(anger-in and anger-out), with physical problems and poor psychologicaladjustment. Combining the clinicalliterature on anger and the findings of the present study, we now know that insecure attachmentpredicts a greater anger experience and greater suppression and outwardexpression of anger.Although further research is needed to identify the factors that determine whether angerisexpressed inwardly or outwardly, we already know that these anger expressions are associatedwith both psychological and physical maladjustment.The results of the present study underscorethe importance of parent-child relations in understandingthe emergence and expression offeelings of anger among at least some adolescents.91In particular, results of the present study demonstrate the importance of fathers foradolescents’ well-being, as well as mothers. Given that separate analyses were employed formother- and father-figure attachments in the present study, we cannot yet determine the relativeinfluence of mother and father attachment figures. However, the pattern of results observed wasconsistent for both mother and father figures, suggesting that both parents may play a role intheadolescents’ anger process. The study of child-father figure attachment has been virtuallyignored within the literature, compared to the number of studies onchild-mother attachments,especially for adolescent populations. Results of the present study add to a small butgrowingnumber of recent studies (Diener, Isabella, & Behunin, 2008; Gomez& McLaren, 2007)demonstrating the significant effects of not only maternal attachmentbut also paternalattachment on children’s peer and academic competence (Diener et al., 2008) andaggressivebehavior (Gomez & McLaren, 2007).Mothers and fathers interact with their children differently, at least in Westerncultures,and there are several differences between mothers and fathers thathave been documented in theliterature (Cox, Owen, Henderson, & Margand, 1992). For example,according to Cox andcolleagues (Cox et al., 1992), whether fathers are highly involved in caretakingor not, they areless likely to hold, tend to show affection toward, smile at and vocalize with theirchild. Olderresearch reported that in families, mothers engage in more caregiving behavior,while fathersserve more as playmates (for review, see Pipp, Easterbrooks, & Brown,1993). More recentresearch in the attachment area, however, emphasizes that mothersand fathers generally showmuch more similarity than differences in their interactions with their child(Zupancic, Podlesek,& Kavcic, 2004). Over the past few decades, the systems of family life havebeen rapidlychanging (Cabrera, Tamis-Lemonda, Bradley, Hofferth, & Lamb, 2000), leadingto92characterizations of both mothers andfathers as co-caregivers. Traditional theories that placemothers as the central influence onchildren’s lives need to be revised in light of suchchanges.The present study, demonstratingthe mediator role of anger experience betweenparentattachment and anger expression,supports theoretical arguments within the emotion literaturethat emotional expressions aremanifestations of internal emotional states (Lewis,2008; Lewis &Michalson, 1983), including anger (e.g., Spielberger,1999). The present findings also underscorethe importance of differentiating essential dimensionsof anger (i.e., intensity of anger and angerexpression) when investigating anger. Infact, distinguishing these two dimensions is consideredkey in the clinical literature (Spielberger,1999).LimitationsThere are several limitations in the present study thatmust be noted. First, a static-nature(single-time point) investigation cannotgenuinely claim a developmental implication. A primaryaim of the present study was toexamine dysfhnctional anger as a predictable correlate ofinsecure attachment based on Bowiby’soriginal hypothesis. Extending this hypothesis, thepresent study also evaluated a path modelin which insecure attachment was specified as apredictor of intensity of anger. Nevertheless, both variableswere concurrently measured in thepresent sample, raising the question of analternative directionality, such as the possibility thatanger may lead to insecure attachment. Another possibility isthat a third variable (e.g.,temperament, personality) leads to both anger and insecureattachment. Consistent with this latterpossibility, some trait theorists (Brussoni, Jang, Livesley, &MacBeth, 2000; Crawford, Livesley,lang, Shaver, Cohen, & Ganiban, 2007;Donnellan, Burt, Levendosky, & Klump, 2008) havesuggested that genetic or temperamentalaspects of anger (i.e., dispositional anger) leads toindividual differences in attachment security,rather than the reverse. To uncover the possible93origins of individual differences inattachment as well as anger experiences, extensivelongitudinal research is necessary to observe developmentalchanges over time, not just over oneor two years, but across childhood andadolescence. Although the stability of attachment frominfancy to adolescence has been empirically documented(Gloger-Tippelt et al., 2002; Hamilton,2000; Main & Cassidy, 1988; Wartner et al.,1994; Waters et al., 2000), future longitudinalresearch may nevertheless benefit from an examinationof the links between attachment, angerand temperament/personality over thelife span.The present study considered felt anger as a mediatorlinking insecure attachment withthe expression of anger (i.e., anger in, anger out).Felt anger was treated as a mediator because itwas considered a necessary prerequisitefor anger expression. Bowlby (1973) describeddysfunctional anger as intensive and persistent anger crossing “the thresholdof intensity” (p.249), suggesting that there is a certainlevel of felt anger that leads to negative consequences.Thus, one might also consider a“threshold” model of anger experience, speculating that anindividual would have to feel a certain minimal levelof anger before such emotion would beexpressed. In addition to the meditational feature ofanger intensity considered in the presentstudy, examining the impact of anger levels onanger expression would be a potentially fruitfulfocus for future research examining the processes of angerexperiences.As another limitation, the present study was not able to investigate apossible integrativemodel of attachment relationships with mother and fatherfigures in the SEM analyses due to theinsufficient sample size. If how maternal and paternal attachmentsinfluence with each othercould be examined in the present study in relation toanger, such evidence would provideimportant information concerning the nature of therelationship between the interactive system ofmother and father figures and the anger process. Thisremains a question for future research.94Further, the present study does not speak tocomplex parent situations or roles of othersignificant adults besides parents. Oursociety is changing and becoming complicated ina way inwhich it often occurs that somechildren may not have the presences of motherand fatherfigures. An expansion from researchon parents to a family system including other familymembers, even to a system of neighbors orcommunity, may provide us with a morecomprehensive understanding of howmultiple attachment experiences predict adolescents’outcomes.Lastly, results of the present study are solely basedon adolescent self-reports, raisingconcerns that the relationships observedare primarily the results of shared method invariance,although the use of SEM minimizes this effect by allowingcorrelations among the error terms.Integration of both adolescents’ perceptions abouttheir parents and parents’ perceptions abouttheir children may be useful to better understanding ofbidirectional relationships betweenadolescents and their parents.Despite these limitations, however, the presentstudy sheds new light on the importantroles of attachment to both mother andfather in relation to adolescents’ anger development,offering insightful knowledge that will be helpful forparents, family members, and educators topromote children’s healthy growththrough emotional maturity.95ReferencesAinsworth, M. S., Blehar, M. C., Waters, B., & Wall, S.(1978). Patterns ofattachment: Apsychological study ofthe strange situation.Hilisdale, NJ: Eribaum.Allen, J. P. (2008). The attachment system inadolescence. In J. Cassidy & P. R. Shaver (Eds.),Handbook ofattachment: Theory,research, and clinical applications (2nd ed., pp. 419-435). New York: The Guildford Press.Allen, J. P., & Land, D. (1999). Attachment inadolescence. In J. Cassidy & P. R. Shaver (Eds.),Handbook ofattachment: Theory, research, and clinical applications (pp. 3 19—335).New York: The Guildford Press.Bartholomew, K. (1990). Avoidance of intimacy: An attachment perspective. Journal ofSocialand Personal Relationships, 7, 147—178.Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles among young adults: A test ofa four-category model. Journal ofPersonality and Social Psychology, 61, 226—244.Bartholomew, K., & Shaver, P. R. (1998). Methods of assessing adult attachment: Do theyconverge? In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory and closerelationships (pp.25-45).New York: The Guilford Press.Biaggio, M. K., & Godwin, W. H. (1987). Relation of depression to anger and hostilityconstructs. Psychological Report 61, 87-90.Blumberg, S., & Izard, C. B. (1985). Affective and cognitive characteristics of depression in 10-and 11-year-old children. Journal ofPersonality and Social Psychology, 49, 194-202.Bosworth, K., Espelage, D. L., & Simon, T. R. (1999). Factors associated with bullying behaviorin middle school students. Journal ofEarly Adolescence, 19, 34 1-362.Bowlby, J. (1973). Attachment and loss, Vol. 2: Separation. New York: Basic Books.96Bowiby, J. (1997). Attachment and loss, Vol. 1: Attachment. London: Pimlico. (Original workpublished 1969)Bowiby, J. (1988). A secure base: Parent-child attachment and healthy human development.London: Routledge.Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measurement of attachment:Anintegrative overview. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theoiyandclose relationships (pp.46-76). New York: The Guilford Press.Brennan, K. A., Shaver, P. R., & Tobey, A. E. (1991). Attachment styles,gender, and parentalproblem drinking. Journal ofSocial and Personal Relationships, 8, 451-466.Bretherton, I. (1985). Attachment theory: Retrospect and prospect. Monographs oftheSocietyfor Research in ChildDevelopment, 50 (1-2, Serial No. 209).Bridewell, W. B., & Chang, E. C. (1997). Distinguishing between, depression, andhostility:Relations to anger-in, anger-out, and anger control. Personality and IndividualDifferences, 22, 587-590.Brody, L. R., & Hall, J. A. (2008). Gender and emotion in context. InM. Lewis, J. M. HavilandJones, & L. F. Barrett (Eds.), Handbook ofemotions (3nd ed.,pp.395-408). New York:The Guildford Press.Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In Bollen,K. A.& Long, J. S. (Eds.), Testing structural equation models (pp.136-162).Newbury Park,CA: Sage PublicationsBrussoni, M. J., Jang, K. L., Livesley, W. J., & MacBeth, T. M. (2000). Geneticandenvironmental influences on adult attachment styles. Personal Relationshzps,7, 283-289.Cabrera, N., Tamis-LeMonda, C. S., Bradley, B., Hofferth, S. & Lamb, M. (2000). Fatherhoodin97the 21st century. ChildDevelopment, 71, 127-136.Calamari, E., & Pini, M. (2003).Dissociative experiences and anger pronenessin late adolescentfemales with different attachmentstyles. Adolescence, 38, 287-303.Cassidy, J. (2008). The nature of thechild’s ties. In J. Cassidy & P. R. Shaver (Eds.), Handbookofattachment: Theory,research, and clinical applications (2nd ed.,pp.3-22). New York:The Guildford Press.Cautin, R. L., & Overholser, J. C. (2001).Assessment of mode of anger expression in adolescentpsychiatric inpatients. Adolescence, 36, 163-170.Chaplin, T. M., (2006). Anger, happiness,and sadness: Associations with depressive symptomsin late adolescence. Journal of YouthandAdolescence 35, 977-986.Chaplin, T. M., Cole, P. M., & Zahn-Waxler,C. (2005). Parental socialization of emotionexpression: Gender differences and relationsto child adjustment. Emotion, 5, 80-88.Clay, D. L., Anderson, W. P., & Dixon, W. A. (1993).Relationship between anger expressionand stress in predicting depression. Journal ofCounseling & Development, 72, 9 1-94.Cohen, J. (1988). Statistical power analysisfor thebehavioral sciences (2nd ed.). Hilisdale, NJ:Eribaum.Conger, R. D., Neppi, T., Kim, K. J.,& Scaramella, L. (2003). Angry and aggressive behavioracross three generations: A prospective, longitudinal studyof parents and children.Journal ofAbnormal Child Psychology, 31, 143-160.Conway, J. M., & Huffcutt, A. I. (2003). A review andevaluation of exploratory factor analysispractices in organizational research.Organizational Research Methods, ó 147-168.Cornell, D. G., Peterson, C. S., & Richards, H. (1999).Anger as a predictor of aggression amongincarcerated adolescents. Journal of Consulting andClinical Psychology, 67 108-115.98Cox, M., Owen, M., Henderson, K.,& Margand, N. (1992). Prediction ofinfant-father andinfant-mother attachment.Developmental Psychology, 28,474-483.Cox, D. L., Stabb, S. D., &Hulgus, J. F. (2000). Anger and depression ingirls and boys: A studyof gender differences. Psychologyof Women Quarterly, 24, 110-112.Crawford, T. N., Livesley, W. J., Jang,K. L., Shaver, P. R., Cohen, P., &Ganiban, J. (2007).Insecure attachment andpersonality disorder: A twin study of adults. EuropeanJournalofPersonality, 21,19 1-208.Diener, M. L., Isabella, R. A.,Behunin, M. G., & Wong, M. S. (2008). Attachment tomotherand father during middlechildhood: Associations with child gender,grade, andcompetence. Social Development,1 7 84-101.Donnellan, M. B., Burt, S. A., Levendosky,A. A., & Kiump, K. L. (2008). Genes, personality,and attachment in adults: Amultivariate behavioral genetic analysis. PersonalityandSocial Psychology Bulletin, 34, 3-16.Fabrigar, L. R., Wegener, D. T., MacCallum,R. C., & Strahan, B. J. (1999). Evaluating the useof exploratory factoranalysis in psychological research. Psychological Methods, 4,272-299.Farmer, T. J. (2002). The experienceof major depression: Adolescents’ perspectives. IssuesinMental Health Nursing, 23, 567-585.Field, T. (2002). Violence and touchdeprivation in adolescents. Adolescence, 37,735-749.Fischer, P. C., Smith, R. J., Leonard, E.,Fugua, D. R., Campbell, J. L., & Masters, M.A. (1993).Sex differences on affectivedimensions: Continuing examination.Journal ofCounseling& Development, 71, 440-443.Floyd, F. J., & Widaman, K.F. (1995). Factor analysis in the development and refinementof99clinical assessment instruments. Psychological Assessment, 7, 286-299.Ford, J. K., MacCallum, R. C., & Tait, M. (1986). The applicationof exploratory factor analysisin applied psychology: A critical reviewand analysis. Personnel Psychology, 39, 291-314.Fraley, R. S., & Wailer, N. G. (1998). Adultattachment patterns: A test of the typological model.In J. A. Simpson & W. S. Rholes (Eds.),Attachment theory and close relationships(pp.77-i 14). New York: The Guilford Press.Freud, S. (1963). Mourning and melancholia.In S. Freud (Ed.), General psychology theory(pp.1953-1974). New York: Collier Books. (Original work published 1917)Gerbing, D. W., & Anderson, J. C. (1993).Monte Carlo evaluations of goodness-of-fit indicesfor structural Equation models. In Boiien, K. A. & Long, J. S.(Eds.), Testing structuralequation models (pp.40-65). Newbury Park, CA: Sage PublicationsGloger-Tippelt, G., Gomille, B., Koenig, L., & Vetter, J. (2002). Attachmentrepresentations in6-year-old: Related longitudinally to the quality of attachment in infancyand mothers’attachment representations. Attachment and Human Development, 4, 318-339.Golman, L., & Haaga, D. A. F. (1995). Depression and the experience andexpression of anger inmarital and other relationships. Journal ofNervous and Mental Disease, 183, 505-509.Gomez, R., & McLaren, S. (2007). The inter-relations of mother andfather attachment, self-esteem and aggression during late adolescence. Aggressive Behavior, 33, 160-169.Gorsuch, R. L. (1997). Exploratory Factor Analysis: Its role initem analysis. Journal ofPersonality Assessment, 68, 532-5 60.Hair, J. E., Black, W. C., Babin, B. J., Anderson, R. E., &Tatham, R. L. (2006). Multivariatedata analysis (6th ed.). Upper Saddle River, NJ: Pearson Prentice Hall.100Hamilton, C. E. (2000). Continuity and discontinuity of attachment from infancy throughadolescence. Child Development, 71, 690-694.Harburg, B., Blakelock, E. H., & Roeper, P. J. (1979). Resentful and reflective coping witharbitrary authority and blood pressure: Detroit. Psychosomatic Medicine, 41, 189-202.Harter, S. (1990). Causes, correlates, and the functional role of global self-worth: A life-spanperspective. In R. J. Sternberg & J. Kolligian, Jr. (Eds.), Competence considered (pp.6’7-97). New Have, CT: Yale University Press.Hattie, J. (1985). Methodology review: Assessing unidimensionality of tests and items. AppliedPsychological Measurement, 9, 139-174.Hazan, C., & Shaver, P. (1987). Romantic love conceptualized as an attachment process.Journal ofPersonality and Social Psychology, 52, 511-524.Helfritz, L. E., & Stanford, M. S. (2006). Personality and psychopathology in an impulsiveaggressive college sample. Aggressive Behavior, 32, 28-37.Hesse, B. (1999). The Adult Attachment Interview: Historical and current perspectives. In J.Cassidy & P. R. Shaver (Eds.), Handbook ofattachment: Theory, research, and clinicalapplications (pp. 393-433). New York: The Guildford Press.Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis:Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.Hubbard, J. A. (2001). Emotion expression processes in children’s peer interaction: The role ofpeer rejection, aggression, and gender. ChildDevelopment, 72, 1426-1438.Jacobs, A. G., Phelps, M., & Rohrs, B. (1989). Assessment of anger expression in children: Thepediatric anger expression scale. Personality and Individual Differences, 10, 59-65Kinsfogel, K. M., & Grych, J. H. (2004). Interparental conflict and adolescent dating101relationships: Integrating cognitive, emotional,and peer influences. Journal ofFamilyPsychology, 18, 505-515.Kline, R. B. (2005). Princzles andpractice ofstructural equation modeling (2nd ed.). NewYork: The Guilford Press.Kobak, R. R., Cole, H. E., Ferenz-Gillies, R.,& Fleming, W. S. (1993). Attachment and emotionregulation during mother-teen problemsolving: A control theory analysis. ChildDevelopment, 64, 23 1-245.Kobak, R. R., & Sceery, A. (1988).Attachment in late adolescence: Working models, affectregulation, and representations of self and others. Child Development59, 135-146.Kooper, B.A. (1993). Role of gender, sex role identity,and type: A behavior in anger expressionand mental health functioning.Journal of Counseling Psychology 40, 232—237.Kopper, V. A., & Epperson, D. L. (1996). The experienceand Expression of Anger:Relationships with gender, gender role socialization,depression, and mental healthfunctioning. Journal of Counseling Psychology, 43, 158-165.Lemerise, E. A., & Dodge, K. A. (20&8). Thedevelopment of anger and hostile interactions. InM. Lewis, J. M. Haviland-Jones, & L.F. Barrett (Eds.), Handbook ofemotions (3rd ed.,pp.730-74 1). New York: The Guildford Press.Lemus, R. L., & Johnson, F. A. (2008).Relationship of Latino gang membership to angerexpression, bullying, ethnic identity, andself-esteem. Journal ofGang Research, 16 13-32.Lewis, M. (2008). The emergence of humanemotions. In M. Lewis, J. M. Haviland-Jones, & L.F. Barrett (Eds.), Handbook ofemotions(3rd ed.,pp.304-3 19). New York: TheGuildford Press.102Lewis, M., & Michalson, L. (1983).Children’s emotions and moods: Developmentaltheory andmeasurement. New York:Plenum Press.MacKinnon, D. P., Lockwood, C. M.,Hoffman, J. M., West, S. G., & Sheets, V.(2002). Acomparison of methods to testmediation and other intervening variableeffects.Psychological Methods, 7, 83—104.Main, M., & Cassidy, J. (1988).Categories of response to reunion with theparent at age 6:Predictable from infant attachmentclassifications and stable over a 1-month period.Developmental Psychology, 24,415-426.Main, M., Kaplan, K., & Cassidy,J. (1985). Security in infancy, childhood andadulthood: Amove to the level of representation. In I.Bretherton & E. Waters (Eds.), Growing pointsof attachment theory and research.Monographs ofthe Societyfor Research in ChildDevelopment, 50 (1-2, Serial No.209).Main, M., & Solomon, J. (1990).Procedures for identifying infants as disorganized/disorientedduring the Ainsworth strangesituation. In M. T. Greenberg, D. Cicchetti, & E.M.Cummings (Eds.), Attachment in the preschoolyears: Theory, research, and intervention(pp. 121-160). Chicago:The University of Chicago Press.McGee, R., Silva, P. A., & Williams,S. M. (1983). Parents’ and teachers’ perceptions ofbehavior problems in seen year oldchildren, Exceptional Child, 30, 151-161.McKay, S., & Moretti, M. M.(2001, August). Comprehensive Adolescent-ParentAttachmentInventory (CAPAI): Preliminary validation. Posterpresented at the AmericanPsychological Association, San Francisco,CA.McWhirter, J. J., McWhirter,B. T., McWhirter, E. H., & McWhirter, R. J. (2003). At-riskyouth:103A comprehensiveresponse: For counselors, teachers,psychologists, and humanserviceprofessionals (3rd ed.). Belmont,CA: Brooks/Cole-ThomsonLearning.Meesters, C., & Muris, P.(2002). Attachment style andself-reported aggression. PsychologicalReports, 90, 23 1-235.Mikulincer, M. (1998).Adult attachment style andindividual differences in functionalversusdysfunctional experiences ofanger. Journal ofPersonality and SocialPsychology, 74,513-524.Miles, J., & Shevlin, M.(2001). Applying regression &correlation: A guidefor studentsandresearchers. Thousand Oaks,CA: Sage Publications.Moreno, J. K., Fuhriman, A.,& Selby, M. J. (1993). Measurement ofhostility, anger, anddepression in depressed and nondepressedsubjects. Journal ofPersonality Assessment6], 511-523.Moretti, M. M., McKay, S., &Holland, R. (2000). The ComprehensiveAdolescent-ParentAttachment Inventory. Unpublishedprotocol.Morizot, J., Ainsworth, A. T., &Reise,S. P. (2007). Toward modernpsychometrics: Applicationof item response theory modelsin personality research. In R. W. Robins,R. C. Fraley, &R. F. Krueger (Eds.),Handbook ofresearch methods in personalitypsychology (pp. 407-423). New York: The GuilfordPress.Muris, P., Meesters, C., Morren,M., & Moorman, L. (2004). Angerand hostility in adolescents:Relationships with self-reportedattachment style and perceived parentalrearing styles.Journal ofPsychosomatic Research, 57 257-264.Newman, J. L., Gray, B. A., & Fuqua,D. R. (1999). Sex differences in therelationship of angerand depression: An empiricalstudy. Journal of Counseling & Development,77, 198-203.104Nandakumar, R. (1993).Assessing essential unidimensionalityof real data. AppliedPsychological Measurement, 17,29-3 8.Nandakumar, R., & Ackerman, T.(2004). Test modeling. In D. Kaplan (Ed.),The Sagehandbook ofquantitative methodologyforthe social science(pp.93-105). ThousandOaks, CA: Sage Publications.Nunnally, J., & Bernstein, I. (1994).Psychometric theory. New York: SageMcGraw-Hill.Pipher, M. (1994). RevivingOphelia: Saving the selves ofadolescent girls. NewYork:Ballantine,Pipp, S., Easterbrooks, M.A., & Brown, S. R. (1993). Attachmentstatus and complexity ofinfants’ self- and other-knowledge whentested with mother and father. SocialDevelopment, 2, 1-14.Riley, W. T., Treiber, E. A., & Woods,M.G. (1989). Anger and hostility in depression.JournalofNervous and MentalDisease, 177, 668-674.Robbins, P. R., & Tanck, R. H. (1997).Anger and depressed affect: Interindividual andintraindividula perspectives. Journal ofPsychology, 131, 489-500.Rubin, I. R. (1969). The angry book.New York: Collier.Ryan, L. G., Miller-Loessi, K., & Nieri,T. (2007). Relationship with adults as predictorsofsubstance use, gang involvement, and threatsto safety among disadvantaged urbanhighschool adolescents. Journal ofCommunityPsychology, 35, 1053-1071.Seidlitz, L., Fujita, F., & Duberstain,P. R. (2000). Emotional experience over time and selfreported depressive symptoms. Personalityand Individual c4fferences, 28, 447-460.Sharkin, B. S. (1993). Anger and gender:Theory, research, and implications. Journal ofCounseling & Development, 71, 386-389.105Singer, J. L. (1995).Repression and dissociation:Implicationsfor personality theoiy,psychopathology, andhealth. Chicago: University ofChicago Press.Sperberg, E. D., & Stabb, S.D. (1998). Depression in womanas related to anger and mutualityin relationships. Psychology of WomanQuarterly, 22, 223-238.Spielberger, C. D. (1988).Manualfor the State-TraitAnger Expression Inventory (STAXI).Odessa, FL: PsychologicalAssessment Resources.Spielberger, C. D. (1999).Manualfor the State-Trait Anger ExpressionInventory (STAXI-2).Odessa, FL: PsychologicalAssessment Resources.Spielberger, C. D., Jacobs, G.,Russell, S., & Crane, R. (1983). Assessmentof anger: The State-Trait Anger Scale. In J. N.Butcher & C. D. Spielberger (Eds.), Advancesin personalityassessment (Vol. 3,pp.112-134). Hillsdale, N. J.: LawrenceEribaum Associates.Spielberger, C. D., Johnson, E. H.,Russell, S. F., Crane, R. J., Jacobs, G. A., &Worden, T. I.(1985). The experience andexpression of anger: Construction and validationof an angerexpression scale. In M. A. Chesney & R.H. Rosenman (Eds.), Anger and hostility incardiovascular and behavioraldisorders (pp.5-30). New York:Hemisphere PublishingCorporation.Steiger, A. R. (2003). Preliminaryvalidation ofthe Comprehensive Adolescent-ParentAttachment Inventory. Unpublished master’sthesis.Stout, W. F. (1987). A nonparametricapproach for assessing latent traitunidimensionality.Psychometrika, 52, 589-6 17.Stout, W. F. (1990). A newitem response theory modeling approachwith applications tounidimensionality assessment andability estimation. Psychometrika, 55,293-325.Swan, S. C., Gambone, L. J.,Fields, A. M., Sullivan, T. P., & Snow, D.L. (2005). Women who106use violence in intimate relationships: Therole of anger, victimization, and symptoms ofposttraumatic stress and depression. Violence and Victims, 20,267-285.Tabachnick, B. G., & Fidell, L. S. (2001). Usingmultivariate statistics (4th ed.). NeedhamHeights, MA: Allyn & Bacon.Thomas, D. R., Hughes, E., & Zumbo,B. D. (1998). On variable importance in linear regression.Social Indicators Research: An International andInterdisciplinary Journal ofQuality-ofLfe Measurement, 45, 253-275.Thomas, D. R., & Zumbo, B. D. (1996). Using a measure ofvariable importance to investigatethe standardization of discriminant coefficients. Journal ofEducational & BehavioralStatistics, 21, 110-130.Thompson, R. A. (1997). Sensitivity and security: New lessons to ponder.ChildDevelopment,68, 595-597.Thompson, R. A. (1999). Early attachment and later development. In J. Cassidy & P. R.Shaver(Eds.), Handbook ofattachment (pp. 265-286). New York: The GuildfordPress.Troisi, A., & D’Argenio, A. (2004). The relationshipbetween anger and depression in a clinicalsample of young men: The role of insecure attachment.Journal ofAffective Disorders,79, 269-272.Underwood, M. K. (2003). Social aggression among girls. New York: Guilford Press.van IJzendoorn, M. H. (1995). Adult attachment representations,parental responsiveness, andinfant attachment: A meta-analysis on the predictive validity of the AdultAttachmentInterview. Psychological Bulletin, 117, 3 87-403.van IJzendoorn, M. H., & De Wolff, M. 5. (1997).In search of the absent father — Meta-analyses107of infant-father attachment:A rejoinder to our discussants.ChildDevelopment, 68, 604-609.Venable, V. L., Carison, C. R., &Wilson, J. (2001). The role of anger and depressioninrecurrent headache. Headache,41, 2 1-30.Wartner, U. G., Grossmann, K.,Fremmer-Bombik, E., & Suess, G. (1994). Attachmentpatternsat age six in South Germany:Predictability from infancy and implications forpreschoolbehavior. Child Development, 65,1014-1027.Waters, E., Merrick, S., Treboux,D., Crowell, J., & Albersheim, L. (2000). Attachmentsecurityin infancy and early adulthood: A twenty-yearlongitudinal study. Child Development71,684-689.Whiteside, S. P., & Abramowitz,J. S. (2005). The expression of anger andits relationship tosymptoms and cognitions in obsessive-compulsivedisorder. Depression and Anxiety, 21,106-111.Zaitsoff, S. L., Geller, J., & Srikameswaren, S.(2002). Silencing the self and suppressed anger:Relationship to eating disorder symptoms inadolescent females. European EatingDisorders Review, 10, 51-60.Zeman, J., Shipman, K., Suveg, C. (2002). Angerand sadness regulation: Predictions tointernalizing and externalizing symptomsin children. Journal ofClinical Child andAdolescent Psychology, 31, 393-398.Zupancic, M., Podlesek, A., & Kavcic,T. (2004). Parental child-care practicesof Slovenianpreschoolers’ mothers and fathers: The Family EnvironmentQuestionnaire. Horizons ofPsychology, 13, 7-26.108Appendix A -1Dear Parent(s) or Guardian(s),We are writing to ask permissionfor your son/daughter to take part in a research project entitled,“Relationships and Anger Among Teens” atyour son/daughter’s school. In this project, we are trying tolearn more about how teens’ experiencesof anger — how much they feel angry and how they expressthatanger (acting out towards others, orkeeping it inside) — are linked to the quality of their relationships withparents and friends. All students in grades8-12 are invited to participate, but first they need your consentto do so. To help you decide, we describe the project foryou here.Study Description: Students will be askedto fill out questionnaires in one group session (approximately45 minutes in length) during class time. Thequestionnaires will ask students: a) about their backgroundinformation (grade, gender, age), b) their feelingsabout how well they feel they get along with bothfriends and family, c) how much and how oftenthey feel angry at things, and d) their ability to regulateand express emotions such as anger.Who Participates: Only students who receiveparent permission will be asked to take part in our projectand students themselves will be askedif they wish to participate. Your son/daughter’s participation isvoluntary and students can withdraw fromthe study at any time without any consequences. Whether ornot a student takes part in this project does not affect theirschoolwork in any way. Students who do notparticipate will be given a classroom activity to complete (decided by teachers)such as reading during thetesting.Confidentiality: All of the information obtained from individual students in thisproject is consideredstrictly confidential and will only be seen by the researchers.All reports of the findings of this project willbe at the level of group findings, not individuals. No nameswill appear on any of the questionnaires.Instead, numbers will be given to each student.Consent: Please indicate on the next page if you give permissionfor your son/daughter to participate ornot. Your son/daughter should then return the form tohis or her teacher by_____________.Please returnthe form even if you do not want your son/daughter to participateso that we know you received ourrequest. You can keep this letter and the top portion of theconsent form for your records. All studentswho return parent/guardian and student consent forms (indicating“yes” or “no”) will have the opportunityto win a $25.00 bookstore gift certificate(one award per class).Contact: We would be very pleased if your son or daughtertakes part in our study and we hope that youwill give her or him permission to do so. If you have any questions,please feel free to call Chiaki Konishi(604-827-21Q4)8or Dr. Shelley Hymel (604-822-6022). If you have any questions about yourson/daughter’s treatment or rights as a research subject, you may contactthe Research SubjectInformation Line in the UBC Office of Research Services at (604) 822-8598.Thank you very much foryour time and consideration of this request.Sincerely,Shelley Hymel, Department Head & Professor Chiaki Konishi,M.A., Ph.D. Student‘PLEASE KEEP THIS LETTER FOR YOUR RECORDS***This research is being conducted in order to fulfill the dissertationrequirements for a Ph.D. degree in thedepartment of Educational and Counselling Psychology, and SpecialEducation at the University of British Columbia.109PARENTIGUARDIAN CONSENT FORMStudy Title: “Relationships andAnger Among Teens”Principal Investigator: Shelley Hymel, Ph.D.Department Head & ProfessorUniversity of British ColumbiaCo-Investigator: Chiaki Konishi, M.A.University of British ColumbiaConsent:I have read and understood the information presented aboutthe study entitled, “Relationships and AngerAmong Teens”.I understand that my son/daughter’s participation in the studyis entirely voluntary, and that he/she maywithdraw from the study at any time without any consequencesor impact on his/her class standing orschoolwork.I have received a copy of thisconsent form for my own records.My decision regarding my son/daughter’sparticipation in the study is indicated below (please check one):_______YES, I give permission for my son/daughter to participate in this study.NO, I do not give permission for my son/daughter to participate in this study.***PLEASE KEEP THIS PORTION FOR YOUR RECORDS******PLEASE RETURN THE BOTTOM HALF OF THE FORM TO THE SCHOOL***PARENTIGUARDIAN CONSENT FORMConsent:I have read and understood the information presented about the studyentitled “Relationships and AngerAmong Teens”.I understand that my son/daughter’s participation in the study is entirely voluntary, andthat he/she maywithdraw from the study at any time without any consequences or impact on his/her classstanding orschoolwork.I have received a copy of this consent form for my own records.My decision regarding my son/daughter’s participation in the studyis indicated below (please check one):YES, I give permission for my son/daughter to participate in this study.NO, I do not give permission for my son/daughter to participate in this study.Son/daughter’s Name (please print):___________________________________________Son/daughter’s Grade:_____________________________Son/daughter’s Birth Date:___________________________Parent/Guardian Signature:Date:110Appendix A-2STUDENT ASSENT FORMDear Students,You are invited to be partof a research project thatwill take place at your school. In thisproject, we are trying tolearn more about how teens’experiences with anger are linked tothe quality of their relationshipswith parents and friends. All studentsin grades 8-12 areinvited to participate.If you take part in this project,you will fill out questionnaires ina group session (about 45minutes), held during regularclass time. In the questionnaires,we will first ask some questionsabout you (your grade, if you are aboy or girl, your age). We will alsoask about how you feelabout the important peoplein your life — friends and family.You will also be asked about yourexperiences with anger —how often you feel angry and howyou express your anger.If you take part in our study,it does not affect your schoolworkor your grades in any way. Youcan choose not to be in thisproject now or at any time; that’sokay. Students who do not want tobe part of our project willbe asked to work on otherschool work that your teacher will assignwhile the other students arefilling out the questionnaires. We hope thatwe can use what we learnin this project to better understandthe challenges that teens face today.All of the information yougive us on our questionnaires isconfidential. You will not put yourname on any of the questionnaires.Also, your answers will not beshown to your teachers,parents, students, or anyother persons in the school. THISIS NOT A TEST. There are norightor wrong answers — justwhat you think. So it is very important thatyou answer all of thequestions as honestly as you can.If you want to take part inour project, please fill out the formon the next page. Thank you verymuch for your help.Sincerely,Shelley Hymel, Ph.D. andChiaki Konishi, M.A.Please see other side111STUDENT ASSENT FORMI have read and understood thedescription of the study, “Relationshipsand AngerAmong Teens.”I understand that it is my decision tobe part of this project or not, and that I can decidenot to take part at any time withoutany problem. I also understand that being in thisproject will not affect my schoolwork at all.Please check below, if you chooseto participate in this project:_____Yes, I agree to participate.Name (Please print):Signature:Grade:Date:112Appendix BEthics ApprovalThe University of British ColumbiaOffice of Research ServicesBehavioural Research Ethics BoardSuite 102, 619Q Agronomy Road, Vancouvei B.C. V6T 1Z3CERTIFICATE OF APPROVAL - FULL BOARDPRINCIPAL INVESTIGATOR: INSTITUTION I DEPARTMENT:UBC BREB NUMBER:UBClEducation/Educational &Shelley Hymel Counselling Psychology, and Special H07-00663EducationINSTITUTION(S) WHERE RESEARCH WILL BE CARRIED OUT:Institution I SiteN/A N/AIther locations where the research will be conducted:‘ublic schools in the Lower Mailand, including Vancouver, West Vancouver and Coquitlam school districtsCO-INVESTIGATOR(S):(Thiaki KonishiSPONSORING AGENCIES:N/APROJECT TITLE:elationships and Anger Among TeenstEB MEETING DATE: CERTIFICATE EXPIRY DATE:ay 24, 2007 IMay 24, 2008DOCUMENTS INCLUDED IN THIS APPROVAL: IDATE APPROVED:IJune 12,2007ocument Name I Version I DateConsent Forms:Parental Consent:Version 1 May 1, 2007ssent Forms:tudent AssentVersion 2 May 29, 2007uestionnaire. Questionnaire Cover Letter. Tests:luestionnaires (including a cover sheet)Version 1 May 1, 2007etter of Initial Contact:Parental Consent:Version 1 May 1, 2007Other Documents:The application for ethical review and the document(s) listed above havebeen reviewed and the procedures wereound to be acceptable on ethical grounds forresearch involving human subjects.Approval Is issued on behalf of the Behavioural Research Ethics Boardand signed electronically by one of the following:Dr. Peter Suedfeld, ChairDr. Jim Rupert, Associate ChairDr. Arminee Kazanjian, Associate ChairDr. M. Judith Lynam, Associate ChairDr. Laurie Ford, Associate Chair113Appendix C-iABOUT YOUWe are interested in learning about your background. Please answer allof the questions honestly.REMEMBER, ALL OF YOUR ANSWERS WILL REMAIN PRIVATE AND CONFIDENTIAL, AND WILLONLY BE SEEN BY THE RESEARCHERS.1. Are you female or male? (Check one): Female Male2. How old are you?: Years Old3. When were you born?:____________________I____________I__________________(month) (day) (year)4. What grade are you in now? (Check one):9th_____ 10th111h_ 12th5. How do you describe yourself in terms of ethnic or cultural heritage?(Check all that apply)_______First Nations (North American Indian, Metis, Inuit, etc.)African I CaribbeanAsian (Chinese, Japanese, Korean, etc.)________South Asian (East Indian, Indonesian, Pakistani, etc.)European (Anglo, European descent, etc.)Latino (Spanish, Mexican, South American, etc.)Middle Eastern (Arabic, Iranian, Israeli, Persian, Turkish, etc.)Other (If you would describe your ethnic or cultural heritage in someway that is not listed above, please describe your ethnic or heritage on the line below.)Thank you!For the following pages, please be sure to read all of the instructions before starting114Appendix C-2ABOUT MY PARENTSInstructions:Please think aboutparents or caregivers that have played themost importantpart in raising you. Youmost likely live with themnow, but you may be livingsomewhere else and stillhave contact with them.Answer all the questions belowbased on how you feelabout them.Read each sentence andcircle the number to showhow much you agree or disagree.Before you start,1. Circle the mother figure youwill be describing (If you have both,choose the one you think of asyour primary mother figure).A. Biological MotherB. Adopted MotherC. Step-Mother (orFather’s Significant Other)D. Other_________________________2. Circle the father figure youwill be describing (If you have both,choose the one you thinkof as yourprimary father figure).A. Biological FatherB. Adopted FatherC. Step-Father (or Mother’sSignificant Other)D. Other1. I prefer not to showmy mother how I feel deepdown.2. When I’m awayfrom my mother I feel anxiousand afraid.3 I am very comfortablebeing close to my mother4. If I can’t get my mother toshow interest in me, Iget upset or angry.5. I find it difficult to dependon my mother.6. I worry about beingaway from my mother.7 I need a lot of reassurancethat I am loved bymy mother8. I worry that mymother won’t care about me asmuch as I care bout my mother.9 I worry aboutbeing abandoned by my mother10. I don’t feel comfortable openingup to mymother.About MotherDisagree NeutrallAgreeStrongly MixedStrongly1 2 3 4 56 71 2 3 45 6 71 2 3 4 56 71 2 3 45 6 71 2 3 45 6 71 2 3 4 5 671 2 3 4 56 71 2 3 4 5 671 2 3 45 61 2 3 4 56 77115About Mother (continued)Disagree Neutral!AgreeStrongly MixedStrongly11. Just when my motherstarts to get close to meI 1 2 34 5 6 7find myself pulling away.12. I get frustrated when mymother is not around1 2 3 4 5 67as much as I wouldlike.13. I feel comfortable sharingmy private thoughts1 2 3 4 5 67and feelings with my mother.14. I get uncomfortablewhen my mother wants to1 2 3 4 5 67be very close.15. I often wish that my mother’sfeelings for me 1 2 34 5 6 7were as strong as my feelingsare for mymother.16. I feel comfortable depending onmy mother. 1 2 34 5 6 717. When my mother disapprovesof me, I feel really 1 2 34 5 6 7bad about myself.18. I try to avoid getting tooclose to my mother. 1 2 34 5 6 719. I worry a lot about my relationshipwith my 1 2 34 5 6 7mother.20. I tell my mother justabout everything.1 2 3 4 5 6 721. I often want to be reallyclose to my mother and 12 3 4 5 6 7sometimes this makes my mother backaway.22. I want to get close to my mother,but I keep 1 2 3 45 6 7pulling back.23. I resent it when my mother spendstime away 1 2 3 45 6 7from me.24. I usually discuss my problemsand concerns 1 2 3 45 6 7with my mother.25. I find it relatively easy toget close to my mother. 1 2 34 5 6 726. Sometimes I feel that Ihave to force my mother 1 2 3 45 6 7to show that my mother caresabout me.27. I don’t mind asking my motherfor comfort, 1 2 34 5 6 7advice, or help.28. My desire to be very close sometimesscares 1 2 3 45 6 7people away.11629. I worry a fair amountabout losing my mother.30. I turn to mymother for many things, includingcomfort and reassurance.31. I prefer not to be tooclose to my mother.32. I get frustrated if mymother is not availablewhen I need my mother.33. It helps to turn to my motherin times of need.34. I find that my motherdoesn’t want to get asclose as I would like.35. I don’t often worry aboutbeing abandoned.36. I am nervous when mymother gets too close tome.1 2 3 45 6 71 2 3 45 6 71 2 3 4 56 71 2 3 4 56 71 2 3 4 5 6 71 2 3 4 5 671 2 3 4 56 71 2 3 4 5 67About Mother (continued)DisagreeNeutral! AgreeStrongly MixedStronglyNow About Father...I I prefer not to show my father howI feel deepdown2. When I’m away from my fatherI feel anxiousand afraid.3. I am very comfortable beingclose to my father.4. If I can’t get my father to showinterest in me, Iget upset or angry.5. I find it difficult to depend onmy father.6. I worry about being awayfrom my father.7 I need a lot of reassurance that Iam loved bymy father.8. I worry that my father won’t careabout me asmuch as I care bout my father.9 I worry about being abandonedby my father10. I don’t feel comfortable openingup to my father.11. Just when my father starts to get closeto me Ifind myself pulling away.12. I get frustrated when myfather is not around asmuch as I would like.13. I feel comfortable sharing myprivate thoughtsand feelings with my father.14. I get uncomfortable when my father wantsto bevery close.15 I often wish that my fathers feelings forme wereas strong as my feelings are for my father.16. I feel comfortable depending onmy father.17 When my father disapproves of me Ifeel reallybad about myself18. I try to avoid getting too closeto my father.19. I worry a lot about my relationshipwith myfather117About FatherDisagree Neutrai!AgreeStrongly MixedStrongly44444555556667771 2 3 4 5 671 2 3 4 5 671 2 3 671 2 3 671 2 31 2 31 2 31 2 3 4 5 671 2 3 4 56 71 2 3 4 5 671 2 3 4 5 671 2 3 4 5 671 2 3 4 5 671 2 3 4 5 671 2 3 4 56 71 2 3 4 5 671 2 3 4 5 671 2 3 4 5 671 2 3 4 5 67118About Father (continued)Disagree Neutral! AgreeStrongly Mixed Strongly20. I tell my father just about everything.1 2 3 4 5 6 721. I often want to be really close to my father and1 2 3 4 5 6 7sometimes this makes my father back away.22. I want to get close to my father, but I keep1 2 3 4 5 6 7pulling back.23. I resent it when my father spends time away1 2 3 4 5 6 7from me.24. I usually discuss my problems and concerns1 2 3 4 5 6 7with my father.25. I find it relatively easy to get close to my father. 1 2 3 4 5 6726. Sometimes I feel that I have to force my father 1 2 3 4 5 6 7to show that my father cares about me.27. I don’t mind asking my father for comfort, 1 2 34 5 6 7advice, or help.28. My desire to be very close sometimes scares 1 2 34 5 6 7people away.29 I worry a fair amount about losing my father 1 2 3 4 5 6 730. I turn to my father for many things, including 1 2 3 4 5 6 7comfort and reassurance.31. I prefer not to be too close to my father. 1 2 3 4 5 6 732. I get frustrated if my father is not available when 1 2 3 4 5 6 7I need my father.33. It helps to turn to my father in times of need. 1 2 34 5 6 734. I find that my father doesn’t want to get as close 1 2 3 4 56 7as I would like.35. ldon’toftenworryaboutbeing abandoned. 1 2 3 4 5 6 736. I am nervous when my father gets too close to 1 2 3 4 5 6 7me.Appendix C-3STAXI-2 QUESTIONNAIRENote: Due to copyright, the questionnaire items were not presented on thispage.1195(U(U>‘VUI2(U(U>C(UU-IAppendix DScree Plots for ExploratoryFactor Analyses100Factor NumberFactor Number(UD(U>C(UUI(UD(U>C(U0)U-I120(UD(U>C(U0)LII(Ua>C(UDlUI86421 3 5 7 9 11 1315 171 3 5 7 9 11 13 15 17Figure 1. Scree plot of eigenvaluesfronthe EFA of the anxious attachment itemMother figure attachment - Entire samp6Figure 2. Scree plot of eigenvalues Ironthe EFA of the avoidant attachment iterrMother figure attachment - Entire samp43201 2 3 4 5 6 7 8 9 100Factor NumberFigure 3. Scree plot of eigenvaluesfrcthe EFA of the trait anger items: Mothfigure attachment - Entire sample.43201 2 3 4 5 6 7 8Factor NumberFigure 4. Scree plot of eigenvalues frcthe EFA of the anger-in items: Motherfigure attachment - Entire sample.Factor NumberFigure 6. Scree plot of eigenvalues from tiEFA of the anxious attachment items. Fatifigure attachment - Entire sample.1 2 3 4 5 6 7 8Factor NumberFigure 5. Scree plot of eigenvalues frcthe EFA of the anger-out items: Mothfigure attachment - Entire sample.(U(‘5>C‘1)0)UI6(UC1)30)UI20Figure 11. Scree plot of eigenvalues frcthe EFA of the anxious attachment itemMother figure attachment - Boys.5)1)(U>c3(U0)UJ2(U(U>CV0)LiiFigure 12. Scree plot of eigenvalues frcthe EPA of the avoitant attachment itexiMother figure attachment - Boys61211210(a>(U0)ui420Factor NumberFigure 7. Scree plot of eigenvalues from IEFA of the avoidant attachment items:Father figure attachment - Entiresample.1 3 5 7 9 11 13 15171 2 3 4 5 6 7 8 9 100Factor NumberFigure 8. Scree plot of eigenvalues from IEPA of the trait anger items: Father figunattachment - Entire sample.432043201 2 3 4 5 6 7 8Factor NumberFigure 9. Scree plot of eigenvalues from IEFA of the anger-in items: Father figureattachment - Entire sample.1 2 3 4 5 6 7 8Factor NumberFigure 10. Scree plot of eigenvalues frcthe EPA of the anger-out items: Fatherfigure attachment - Entire sample.9876543210.,,,,,,,,,,.,.,.—.-._.__.___.._.1 3 5 7 9 11 13 15 17Factor Number(U(U>C(U0)Lu1 3 5 7 9 11 13 15 17Factor Number1 2 3 4 5 6Factor NumberFigure 15. Scree plot of eigenvalues frcthe EFA of the anger-out items Motherfigure attachment - Boys.1210Cu>Ca>u420—•,•••*,*,*.T1 3 5 7 9 11 13 15 17Factor NumberFigure 17. Scree plot of eigenvalues frothe EFA of the avoidant attachment itenMother figure attachment - Girls.1 2 3 4 5 6 7 8Factor NumberFigure 14. Scree plot of eigenvalues frcthe EFA of the anger-in items: Motherfigure attachment - Boys.12243201 2 3 4 5 6 7 8 9 1054I:0Factor NumberFigure 13. Scree plot of eigenvaluesfrcthe EFA of the trait anger items:Mothefigure attachment - Boys.43a)DCu>Ca)U]076543207 8a)>Ca>a’U]a)(‘3>Ca)a’Lu65a)(‘3>Ca)03U]0Factor NumberFigure 18. Scree plot of eigenvalues frcthe EFA of the trait anger items: Mothefigure attachment - Girls.1 3 5 7 9 11 13 15 17Factor NumberFigure 16. Scree plot of eigenvalues frcthe EPA of the anxious attachment itemMother figure attachment - Girls.L1 2 3 4 5 6 7 8 9 10(U:3(U>(UuJ1234320Factor NumberFigure 20. Scree plot of eigenvalues frcthe EFA of the anger-out items: Motherfigure attachment - Girls.1 2 3 4 5 6 7 8(U:3(U>(UUI(U:3(U>C(UUI43(U:3(U>(UUI01 2 3 4 5 67 8Factor NumberFigure 19. Scree plot of eigenvaluesfrcthe EFA of the anger-in items: Motherfigure attachment - Girls.876(U(U>C(U. 3LII205(U>C(U[Li21 3 5 7 9 11 13 1517Factor NumberFigure 21. Scree plot of eigenvalues fro:the EPA of the anxious attachment itemsFather figure attachment - Boys.1 3 5 7 9 11 13 15 176Factor NumberFigure 22, Scree plot of eigenvalues frothe EFA of the avoidant attachment iterrFather figure attachment - Boys.0432101 2 3 4 5 6 7 8 9 10Factor NumberFigure 23. Scree plot of eigenvalues frcthe EPA of the trait anger items: Fatherfigure attachment - Boys.1 2 3 4 5 6 7 8Factor NumberFigure 24. Scree plot of eigenvalues frcthe EPA of the anger-in items: Fatherfigure attachment -Boys.0)3(U>C4)0)LU4)3(U>C0)0)LUU)3(U>CU)0)LUFigure 29. Scree plot of eigenvalues froitthe EFA of the anger-in items: Father figiattachment - Girls.0)3(U>CU)0)LU01 2 3 4 5 6 7 8 9 10Factor NumberFigure 28. Scree plot of eigenvalues frcthe EFA of the trait anger items: Fatherfigure attachment - Girls.3U)3(U>c2U)0)LU0Figure 30. Scree plot of eigenvalues frcthe EFA of the anger-out items: Fatherfigure attachment - Girls.1241 2 3 4 5 6 7 84320Factor NumberFigure 25. Scree plot of eigenvalues frcthe EFA of the anger-out items: Fatherfigure attachment - Boys.12101 3 5 7 9 11 13 15 17764)53C0)301LU20Factor NumberFigure 26. Scree plot of eigenvalues frcthe EFA of the anxious attachment itemFather figure attachment - Girls.66420L1 3 5 7 9 11 13 15 17Factor NumberFigure 27. Scree plot of eigenvalues frothe EFA of the avoidant attachment itenFather figure attachment - Girls.4321041 2 3 4 5 6 7 8Factor Number1 2 3 4 5 6 7 8Factor NumberAppendixECorrelationMatricesforScaleItemsTable1IntercorrelationsAmongtheAttachmentAnxietyScaleItemsforMotherFigureAttachment:EntireSample(N=775)ItemXlX2X3X4X5X6X7X8X9X10XliX12X13X14X15X16X17X18Xi1.00X2.311.00X3.65.361.00X4.30.41.381.00X5.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:X=Attachmentanxiety.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.tJTable2IntercorrelationsAmongtheAttachmentAvoidanceScaleItemsforMotherFigureAttachment:EntireSample(N=775)ItemViV2V3V4V5V6V7V8V9V10ViiV12Vi3V14V15V16V17V18Vi1.00V2.431.00V3.35.411.00V4.61.49.451.00V5.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:V=Attachmentavoidance.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Table3IntercorrelationsAmongtheIntensityofAngerScaleItemsforMotherFigureAttachment:EntireSampleN=775)ItemTiT2T3T4T5T6T7T8T9T10Ti1.00T2.721.00T3.70.761.00T4.39.41.401.00T5.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:T=Intensityofanger.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Table4IntercorrelationsAmongtheAnger-InScaleItemsforMotherFigureAttachment:EntireSample(N=775)ItemIi12131415161718Ii1.0012-.061.0013.13.331.0014.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:I=Anger-in.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.L’-) 00Table5IntercorrelationsAmongtheAnger-OutScaleItemsforMotherFigureAttachment:EntireSample(N=775)Item0102030405060708011.0002.231.0003.43.111.0004.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:0=Anger-out.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Table6IntercorrelationsAmongtheAttachmentAnxietyScaleItemsforFatherFigureAttachment:EntireSample(N=762)ItemXlX2X3X4X5X6X7X8X9X10XliX12X13X14X15X16X17X18Xl1.00X2.421.00X3.70.451.00X4.30.41.401.00X5.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:X=Attachmentanxiety.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Table7IntercorrelationsAmongtheAttachmentAvoidanceScaleItemsforFatherFigureAttachment:EntireSample(N=762)ItemViV2V3V4V5V6V7V8V9V10ViiV12V13V14V15V16V17V18Vi1.00V2.511.00V3.46.461.00V4.65.52.541.00V5.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:VAttachmentavoidance.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Table8IntercorrelationsAmongtheIntensityofAngerScaleItemsforFatherFigureAttachment:EntireSample(N=762)ItemTiT2T3T4T5T6T7T8T9T10Ti1.00T2.731.00T3.70.751.00T4.39.42.411.00T5.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:T=Intensityofanger.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Table9IntercorrelationsAmongtheAnger-InScaleItemsforFatherFigureAttachment:EntireSample(N=762)ItemIi12131415161718Ii1.0012-.061.0013.11.341.0014.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:I=Anger-in.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Table10IntercorrelationsAmongtheAnger-OutScaleItemsforFatherFigureAttachment:EntireSample(N=762)Item0102030405060708011.0002.241.0003.43.111.0004.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:0=Anger-out.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Table11IntercorrelationsAmongtheAttachmentAnxietyScaleItemsforMotherFigureAttachment:Boys(n=379)andGirls(n396)ItemXlX2X3X4X5X6X7X8X9X10XliX12X13X14X15X16X17X18Xl-.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:X=Attachmentanxiety.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Thecorrelationsbelowthediagonalareforgirlsandthecorrelationsabovethediagonalareforboys.Table12IntercorrelationsAmongtheAttachmentAvoidanceScaleItemsforMotherFigureAttachment:Boys(n=379)andGirls(n=396)ItemViV2V3V4V5V6V7V8V9ybviiV12V13V14V15V16V17V18Vi-.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:V=Attachmentavoidance.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Thecorrelationsbelowthediagonalareforgirlsandthecorrelationsabovethediagonalareforboys.Table13IntercorrelationsAmongtheIntensityofAngerScaleItemsforMotherFigureAttachment:Boys(n=379)andGirls(n=396)(ternTiT2T3T4T5T6T7T8T9T10Ti-.,polychoriccorrelationswereused.Thefirstletterofthevariablenarneindicatesthescaleitbelongsto:T=Intensityofanger.Thenurnbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Thecorrelationsbelowthediagonalareforgirlsandthecorrelationsabovethediagonalareforboys.Table14IntercorrelationsAmongtheAnger-InScaleItemsforMotherFigureAttachment:Boys(n=379)andGirls(n=396)ItemII12131415161718II-.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:I=Anger-in.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Thecorrelationsbelowthediagonalareforgirlsandthecorrelationsabovethediagonalareforboys.00Table15IntercorrelationsAmongtheAnger-OutScaleItemsforMotherFigureAttachment:Boys(n=379)andGirls(n=396)Item010203040506070801-.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:0=Anger-out.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Thecorrelationsbelowthediagonalareforgirlsandthecorrelationsabovethediagonalareforboys.-‘ ScTable16IntercorrelationsAmongtheAttachmentAnxietyScaleItemsforFatherFigureAttachment:Boys(n=369)andGirls(n=393)ItemXlX2X3X4X5X6X7X8X9X10XliX12X13X14X15X16X17X18Xl-.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:X=Attachmentanxiety.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Thecorrelationsbelowthediagonalareforgirlsandthecorrelationsabovethediagonalareforboys.Table17IntercorrelationsAmongtheAttachmentAvoidanceScaleItemsforFatherFigureAttachment:Boys(n=369)andGirls(n=393)ItemViV2V3V4V5V6V7V8V9V10ViiV12V13V14V15V16V17V18Vi-.51.44.6646.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:V=Attachmentavoidance.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Thecorrelationsbelowthediagonalareforgirlsandthecorrelationsabovethediagonalareforboys.Table18IntercorrelationsAmongtheIntensityofAngerScaleItemsforFatherFigureAttachment:Boys(n369)andGirls(11=393)TiT2T3T4T5-. .38 .29 .48 .64 .35 .38 .52.34Item Ti T2 T3 T4 T5 T6 T7 T8 T9 Ti0Note.Giventhecategoricalnatureofthemeasures,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:T=Intensityofanger.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Thecorrelationsbelowthediagonalareforgirlsandthecorrelationsabovethediagonalareforboys.Table19IntercorrelationsAmongtheAnger-InScaleItemsforFatherFigureAttachment:Boys(n=369)andGirls(n=393)ItemIi12131415161718II-.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:I=Anger-in.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Thecorrelationsbelowthediagonalareforgirlsandthecorrelationsabovethediagonalareforboys.Table20IntercorrelationsAmongtheAnger-OutScaleItemsforFatherFigureAttachment:Boys(n=369)andGirls(ii=393)Item010203040506070801-.,polychoriccorrelationswereused.Thefirstletterofthevariablenameindicatesthescaleitbelongsto:0=Anger-out.Thenumbersinthefirstcolumnandinthefirstrowofthetabledenotetheitemnumbersinthescale.Thecorrelationsbelowthediagonalareforgirlsandthecorrelationsabovethediagonalareforboys.


Citation Scheme:


Usage Statistics

Country Views Downloads
United States 8 5
Japan 5 0
India 2 0
Germany 1 9
China 1 0
Canada 1 0
City Views Downloads
Unknown 5 11
Tokyo 5 0
Ashburn 3 0
Washington 1 0
Montreal 1 0
Kansas City 1 0
Beijing 1 0
Huntington 1 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}
Download Stats



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items