Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Mapping the care domain : conceptualization, assessment, and relation to eating disorders Hennig, Karl H. 1999

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1999-463540.pdf [ 16.15MB ]
Metadata
JSON: 831-1.0089280.json
JSON-LD: 831-1.0089280-ld.json
RDF/XML (Pretty): 831-1.0089280-rdf.xml
RDF/JSON: 831-1.0089280-rdf.json
Turtle: 831-1.0089280-turtle.txt
N-Triples: 831-1.0089280-rdf-ntriples.txt
Original Record: 831-1.0089280-source.json
Full Text
831-1.0089280-fulltext.txt
Citation
831-1.0089280.ris

Full Text

MAPPING T H E C A R E DOMAIN: CONCEPTUALIZATION, ASSESSMENT, A N D R E L A T I O N TO EATING DISORDERS by K A R L H . HENNIG B E d . , University of British Columbia, 1982 M.C.S., Regent College, 1994 M.A. , University of British Columbia, 1995 A DISSERTATION SUBMITTED IN PARTIAL F U L F I L L M E N T OF THE REQUIREMENTS FOR THE D E G R E E OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF G R A D U A T E STUDIES Department of Psychology We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A October 1999 © Karl H. Hennig, 1999 In presenting this thesis in partial fulfillment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Psychology The University of British Columbia Vancouver, Canada Date: 13 October 1999 A B S T R A C T The purpose of this dissertation was to undertake a conceptualization and empirical "mapping" of the ethic-of-care domain - often characterized as self-referential and lacking in rigor. The current focus is upon conventional forms of care, involving notions of moral "goodness" as self-silencing and -sacrificial. Employing a "super" circumplex as a prescriptive and descriptive tool, projected item analyses were undertaken as a theoretico-structural clarification of existing scale items, along with a provisional pool of additional items generated as part of Study 1. Based in part on proposed circular criteria, the Conventional Care Scales (CCS) were developed and submitted to a conjoint principal components analysis along with the battery of other care/dependency scale items. An examination of item circular distributions, factor loadings, alpha-contribution plots, and thematic content revealed several factors expressive of two "faces" of conventional care, submissive and ingenuous. These two forms were shown to have unique correlates with measures of adjustment, interpersonal competencies, other factors of the Five-Factor Model, false-self beliefs, and reported distress in narrated accounts of rejected care giving. Gender differences in the association between indices of adjustment and conventional care were also found. The range of conventional care was also extended through the development of scales reflecting other-directed and socially prescribed dimensions of conventional care. The factor structure for the CSS was also replicated in a second sample (Study 2). Participants for Studies 1 and 2 were composed of undergraduate students (N = 302 in both samples) who completed a battery of questionnaires in the first study and the CCS alone in the second study. Anticipating future clinical directions, secondary analyses using structural equation modelling were conducted on an existing data set (N = 92) which included measures of conventional care and perfectionism, along with indices of psychological adjustment and eating disordered attitudes. Results indicated that conventional care, for which there is little research, was more predictive of adjustment and eating disordered attitudes than ' i i i perfectionism, for which there exists a large clinical literature. This research contributes to an understanding of ways in which an ethic of care can "go awry," as well as proposes a research platform upon which the clinical implications of morality and self-ideals can be investigated. These findings speak to both the constraints and prescriptions that can inform a philosophical ethic of care. iv T A B L E OF CONTENTS Abstract i i List of Tables viii List of Figures ix Acknowledgements xi Chapter 1: Introduction 1 Kohlberg and the Justification of Norms 5 Gilligan and the Voice of Care 8 Empirical Investigations 11 Women's Interdependent Self 11 The "Voices" of Care versus Justice 13 Silencing the Self 15 Dependency, Attachment, and Care 20 Future Directions and Finding Common Ground 24 Clinical Applications 30 Summary and Overview 33 Chapter 2: Mapping and Scale Development Studies 37 Introduction 37 Construct Explication 39 Provisional Item Development 40 Method 40 Participants 40 Demographic Information 42 Factor Structure Measures 42 Dependency and Conventional Care Measures 44 Outcome Measures 47 V Results 51 Creating the "Super" Circumplex Space 53 Projecting Outside Variables onto Circular Space 57 Circular Analysis of Scale Items 61 Scale Development and Fidelity to Circular Structure Criteria 69 Conventional Care Scale Development 72 Preliminary Orientation to Scales and their Projections 81 Introduction to the Conjoint Principal Components Analyses 85 Conjoint Principal Components Analysis: Octant HI (Unassured-Submissive) 86 Conjoint Principal Components Analysis: Octant JK (Unassuming-Ingenuous) 101 Further Mapping: Octant NO (Gregarious-Extraverted) 109 Further Mapping: Octant F G (Aloof-Introverted) 113 Beyond the Self-directed Dimension 114 The Super Circumplex as a Nomological Net: Octant D E (Cold-hearted) 120 The Super Circumplex as a Nomological Net: Octant F G (Aloof-Introverted) 124 The Super Circumplex as a Nomological Net: Octant HI (Unassured-Submissive) 125 The Super Circumplex as a Nomological Net: Octant JK (Unassuming-Ingenuous). . • 127 The Super Circumplex as a Nomological Net: Octant NO (Gregarious-Extraverted) 128 Interpersonal Competencies 131 Beyond the Circumplex 136 Gender Differences: Factor Structure Measures 139 Gender Differences: Care/dependency and Outcome Measures 143 Gender Differences: False-self 145 Gender Differences: Motives for Helping 147 vi Regression Analyses Predicting Adjustment 150 Discussion 154 Chapter 3: Conventional Care and Eating Disorders 161 Introduction 161 Eating Disorders: Epidemiology and Characteristics 163 The Thinness Ideal: Why Women? Why Now? 167 Developmental Aspects: Why Young Women? 174 Care, Dependency, and Eating Disorders 177 Perfectionism 179 Method 182 Participants 182 Measures 182 Results 185 Structural Equation Modelling 185 Modelling Adjustment 187 Modelling Weight/shape Concerns 188 Discussion 189 Chapter 4: General Discussion 194 Clinical Implications 200 Differences in Gender or Power/dependency? '. 201 Conventional Care and Differentials of Power 202 Conventional Care and Dependency 204 Implications for a Theory of Authentic Care 207 Limitations and Future Directions 209 Summary 216 Philosophical Excursus 218 References 224 vii Appendix A: Levels of Justice and Care 247 Appendix B: Study 1 Measures 249 Appendix C: Study 2 Conventional Care Scales 278 Appendix D: Conjoint Principal Components Analysis for the Self-directed Care/dependency Scales . . . . ; 283 Appendix E: Conjoint Principal Components Analysis for the Preliminary Anger-in Scale 286 Appendix F: Regression Analyses 287 Appendix G: DSM-IV Eating Disorder Diagnostic Criteria 292 Appendix H: Study 3 Measures 294 Appendix I: Correlations and Descriptive Statistics for the Eating Disorders Analyses 299 LIST OF T A B L E S Table 1: Comparison between Self-silencing Subscales and Measures of Care and Dependency 19 Table 2: Comparison of Scale Circular Statistics for Dependency Measures 67 Table 3: Comparison of Scale Circular Statistics for Measures of Conventional Care . . 68 Table 4: Factor Loadings for Self-directed Conventional Care Scales 75 Table 5: Factor Loadings for the Other-directed and Socially Prescribed Conventional Care Scales 76 Table 6: Descriptive Statistics for the Conventional Care Scales 77 Table 7: Intercorrelations among Care Scales 83 Table 8: Summary of the Conjoint Principal Components Analysis 108 Table 9: Factor Loadings for Non-self-directed Care Scales 118 Table 10: Correlations between Care/Dependency Scales and Interpersonal Competencies 133 Table 11: Correlations between Care Scales and Self-orientations 147 Table 12: Correlations between Care Scales and Care Narratives 149 Table A - l : Kohlberg's Six Stages of Moral Judgment 247 Table A-2: Gilligan's Three Levels of Care 248 Table F - l : Hierarchical Regression Analyses Predicting Adjustment (Depression and Self-esteem) for Octant D E 287 Table F-2: Hierarchical Regression Analyses Predicting Adjustment (Depression and Self-esteem) for Octant HI 288 Table F-3: Hierarchical Regression Analyses Predicting Adjustment (Depression and Self-esteem) for Octant JK 289 Table F-4: Curve Estimation using Regression Analyses for Care/dependency Scales Predicting Adjustment (Depression and Self-esteem) 290 ix LIST OF FIGURES Figure 1: Dependency as an Aspect of Attachment-detachment and Directiveness-receptiveness Dimensions 22 Figure 2: Structural Representation of Agency and Communion 25 Figure 3: Items and Circular Locations for the Interpersonal Adjectives Scales 28 Figure 4: Three Dimensions of Conventional Care and their Respective Scales 41 Figure 5: A Comparison of IASR and IIP-C Scale Items from Octant JK (Unassuming-Ingenuous) within Super Circumplex Space 54 , Figure 6: Structure of the IASR and the IIP-C in "Super" Circumplex Space 56 Figure 7: Projection of an Outside Variable onto an Interpersonal Space 58 Figure 8: Circular Dispersion of Items for the IBT - Demand for Approval (dependency) Subscale 63 Figure 9: Circular Dispersion of Items for the STSS, UCS, and the SPS-SD scales . . . . 64 Figure 10: Circular Dispersion of the Provisional Pool of Self-directed Conventional Care Items 65 Figure 11: Item Contribution to Alpha as a Function of Deviation from the Circular Mean for the IBT - Demand for Approval Subscale 71 Figure 12: Projection of Self-directed Conventional Care Scales onto the Super Circumplex 82 Figure 13: Typical Individual or Group Profile for the IASR or IIP-C Circumplex Data . . 85 Figure 14: Item Contribution to Alpha as a Function of Circumplex Angular Location for the Self-directed Social Perfectionism Scale 88 Figure 15: Item Contribution to Alpha as a Function of Circumplex Angular Location for the External Subscale 93 Figure 16: Item Contribution to Alpha as a Function of Circumplex Angular Location for the Silence Subscale 94 Figure 17: Item Contribution to Alpha as a Function of Circumplex Angular Location for the Care Subscale 104 Figure 18: Item Contribution to Alpha as a Function of Circumplex Angular Location for the Unmitigated Communion Scale 106 Figure 19: Characteristic Attitudes and Behaviors in Circular Space 110 X Figure 20: Super Circumplex Scale Projections for the Other-directed Social Perfectionism, Conventional Care, and Socially Prescribed Conventional Care Scales 117 Figure 21: Personality and Interpersonal Correlates of the Care and Dependency Scales for Octant HI (Unassured-Submissive) 137 Figure 22: Personality and Interpersonal Correlates of the Care and Dependency Scales for Octant JK (Unassuming-Ingenuous) 138 Figure 23: Personality and Interpersonal Correlates of the Care and Dependency Scales for Octant D E (Cold-hearted) 139 Figure 24: IASR Profiles for Women 141 Figure 25: IASR Profiles for Men 142 Figure 26: Scatterplot of Anxious Concern and Depression, with Linear and Quadratic Trend-lines 154 Figure 27: Path Diagram Showing the Structural Model for Perfectionism and Global Self-silencing (Conventional Care) in Predicting Adjustment 187 Figure 28: Path Diagram Showing the Structural Model for Perfectionism and Global Self-silencing (Conventional Care) in Predicting Weight/shape Concerns . . .188 Figure 29: Schematic Summary of the Results of Structural and Substantive Analyses for the Universe of Care Content 196 XI A C K N O W L E D G E M E N T S Without the contributions of a great many wonderful and supportive persons the completion of this degree would not only have been impossible, but never aspired to. I thank Drs. Carl Armerding and Elmer Dyck for their initial encouragement and inspiration, members of my supervisory committee, Drs. Josie Geller, Wolfgang Linden, and Michael Chandler, for their assistance and mentorship. I especially want to thank Dr. Larry Walker for having truly been with me throughout those many challenging years and significant transitionary moments. He has been the very best friend, godfather, mentor, and colleague. I would like very much to thank the British Columbia Health Research Foundation and the Social Sciences and Humanities Research Council of Canada for fellowships supporting my doctoral studies. I also acknowledge the financial support for my dissertation research provided by a British Columbia Health Research Foundation grant and a U B C Humanities and Social Sciences research grant (to Dr. Walker). I had the good fortune of having several hard-working student research assistants. Above all I would like to acknowledge the patience, love, and support of my wife, Mary, without whom the receiving of this degree would have been left as an unfulfilled dream. 1 CHAPTER 1: INTRODUCTION The general purpose of this dissertation was to undertake an empirical investigation -a "mapping" - of the domain of conventional care; best understood as the intersect of both terms, "conventional" and "care." It is the hope that the field of moral psychology might make an important contribution to both personality and clinical assessment through the theoretical and empirical selection/development of a final set of measures most soundly representing the domain of conventional care. The term "conventional," while potentially evoking a range of meanings, is here intended to make explicit reference to the work of Lawrence Kohlberg (e.g., 1981, 1986) which over the past several decades has dominated the field of moral psychology. Reportedly arising "in part as a response to the Holocaust" (Kohlberg, 1981, p. 470), Kohlberg's efforts were of a piece with social psychology's studies in compliance (e.g., Milgram, 1963) and work of the Frankfurt School (Adorno, Frenkel-Brunswik, Levinson, & Sanford, 1950), all of which sought an evaluative base for examining the validity claims of social norms - thus offering a critical theory of conventional society. Kohlberg, in his now classic 1971 article, "From Is to Ought: How to Commit the Naturalistic Fallacy and Get Away with It" proposed that moral validity claims could be justified by pointing to their place in the unfolding course of human development. Later stages of development were said to provide an increasingly adequate basis for moral decision-making. In the current context, the term "conventional" takes its meaning from the second of three levels within Kohlberg's stage theory, first emerging around early adolescence. Kohlberg referred to this developmental level as a "good boy/girl" morality, characterized by a mutual understanding of relationships perspective. It was with respect to the conventional level of development that Kohlberg framed a third, post-conventional level of moral reasoning, anchored by the Platonic conception that "virtue is ultimately one, not many, and ... the name of this ideal form is justice" (Kohlberg, 1981, p. 30). A fuller explication of Kohlberg's levels and stages of moral development is found in Appendix A. 2 Kohlberg's stages of moral development were specifically stages of justice reasoning. The second term in the phrase "conventional care" takes its reference from Carol Gilligan's so called "ethic-of-care." Initiating what came to be called the Gilligan-Kohlberg debate, Gilligan's seminal 1982 book, In a Different Voice, was levelled not only against justice as the final arbiter of the moral good, but also against the individualistic conception of selfhood which Kohlberg's theoretical framework was thought to assume. Discussions in the area initiated by Gilligan have now rippled far and beyond their initial point of entry into the theoretical literature on moral development. Her "voice" metaphor has effectively become a stereotypic catch phrase within certain feminist circles, along with her supposedly more interpersonal conception of selfhood. The ethic of care became both a source of empowerment (by affirming it as a valid morality distinct from justice concerns), as well as a target for (feminist) social analysis in examining "conventions of goodness where the good woman is 'selfless' in her devotion to meeting others' needs" (Gilligan, 1990, p. 9). From the perspective of Gilligan's initially proposed developmental schema and its three levels of care, the current dissertation undertakes an examination of her Level 2 care. See Appendix A for a fuller description of Gilligan's three levels of care. Skoe's sketch of Gilligan's Level 2 "conventions of goodness," or what is here referred to as conventional care, will serve as an initial working definition: This perspective is characterized by a strong emphasis on responsibility and maternal/paternal morality that seeks to provide care for the dependent and unequal. "Good" is equated with self-sacrificing care for others. The person adopts societal values, and conventionally-defined goodness becomes the primary concern because survival is now seen to depend on the acceptance of others. "Right" is defined by others and responsibility for defining it rests with them. The person has a strong need for security and avoids taking responsibility for choices made. S/he feels responsible for the actions of others whereas others are responsible for the choices she or he makes. The strength in this position lies in its capacity for caring; the limitation lies in the prohibition of self-assertion. Conflict arises specifically over the issue of hurting and others are helped or protected often at the expense of self-assertion, (p. 13) While both Kohlberg and Gilligan began with the notion of the "conventional" as 3 potentially oppressive and something to be transcended, providing a moral vantage point from which it might be critiqued, they did so by moving in different directions. Kohlberg was interested in universal moral judgments, whereas Gilligan moved toward the more contextual and particular, especially towards the party or identity politics of women's unique moral concerns. Gilligan speaks most frequently about two aspects of care, avoiding harm and pursuing care; or what will here be referred to as the two faces of conventional care, "submissive care" and "ingenuous care," respectively. Rather than view the two as mutually defining or reflecting a singular global self-silencing construct (Jack, 1991), my emphasis is upon an exploration of the construct's multifaceted nature. Five specific goals guided this dissertation. The first purpose was to provide a circular structure examination and clarification of the conventional care construct. The universe of semantic content was composed of existing scales, as well as an additional provisional pool of items generated as a part of this dissertation. What is meant by a circular structure examination will require some explanation in what follows, but the bulk of its detailing will be undertaken in Chapter 2. The construct of dependency is of particular comparative interest, on the expectation that it should share much of the same semantic space as conventional care. Following Pincus and Gurtman (1995), conventional care and dependency scales were examined within a circular "super" factor space circumscribed by the two axes of dominance-submissiveness and nurturance-cold heartedness. The circular factor space in which scales were plotted was employed, both in its descriptive role as well as its prescriptive role: descriptive, in that circular space can function as a nomological net in which to undertake the task of construct validation. Its prescriptive role is operative in two ways: (a) by prescribing a location within circular space where mature care can be found, linking this with a more philosophical discussion undertaken in the final chapter of this dissertation; and (b) a more narrowly psychometric 4 construal in prescribing a set of circular criteria for use in test development. Second, the process of structural clarification, or "mapping," of the conventional care domain was continued through the development of additional scales from the provisional • pool of items, entitled the Conventional Care Scales. The development of these scales made explicit use of the proposed circular criteria - in addition to the more usual substantive item-total correlations and factor loadings - in their final item selection. Third, the process of construct validation was extended to include the familiar full Five-Factor Model and a variety of outcome constructs: interpersonal competencies, depression, self-esteem, false-self attitudes, and the narrated experience of distress as a rejected caregiver. Finally, and in line with the recent shift in moral psychology (Walker & Hennig, 1997) toward an examination of real-life over abstract hypothetical dilemmas, this dissertation was intended to locate itself within an applied behavioral context by examining the role of conventional care in predicting problems for which women are primarily at risk, eating disorders. This investigation takes the form of some preliminary structural equation modelling of an existing data base. The purpose of this preliminary investigation was to examine, within a largely eating-disordered sample, the comparative relation between conventional care (for which there is a sparse literature) and perfectionism (for which there exists an extensive literature) in predicting adjustment and weight/shape concerns. By way of an overview to what follows, the notion of a Kohlbergian convention will be examined in more detail, followed by a similar expansion on the work of Gilligan and her ethic of care. The sketch of Gilligan's ideas will be followed by an empirical literature review reporting evidence in favor of her proposal: (a) that women possess a more interdependent, less bounded and individuated sense of self than men; and (b) that women construe moral dilemmas more from within a care orientation than a justice orientation in comparison to men. On Gilligan's (1982, p. 2) interpretation, it is precisely this interdependence of the female self that accounts for women's interpersonal sensitivity to 5 issues of responsiveness and care, those very elements said to be absent from male-individuated construals of both the self and the moral domain. Receiving the greatest emphasis in the present investigation is Jack's (1991) empirical operationalization of Gilligan's Level 2 ethic of care, or what is here referred to as "conventional care." Jack's Silencing the Self Scale is the most broadly employed scale of Gilligan's Level 2 ethic of care, and will be discussed in more detail, although in Study 1 several other care-like scales will be examined as well. The expectation that will be fleshed out is that the ever-burgeoning development of "care" scales might be usefully clarified and constrained by considering their relation both to each other, and to the dependency literature. It is hoped that this analysis might mitigate the charge of self-insularity that has been leveled against Gilligan and her colleagues. The use of a structural model, such as that employed in the current investigation, is particularly suggestive, allowing for the comparison and contrast of scales within a well-recognized semantic space. Kohlberg and the Justification of Norms Norms involve shared behavioral expectations. Let's suppose that person A does something that person B considers wrong in some broad moral sense. Person B may say, "Why did you do that?," or, if person A is found in the midst of doing what they shouldn't have, "What do you think you're doing?" In response, a person could: (a) point to mitigating circumstances (e.g., "Normally I wouldn't do that but...."); or (b) claim ignorance or offer a counter-justification (e.g., "What's wrong with that?"). Note that the question itself calls for a response, which the accused feels some force to rejoin. Some give-and-take of reasons and justifications ought to naturally ensue. These sorts of interactions take place regularly. "Why have you come home late? The supper is cold and you didn't phone." Asserting the expectation and the felt need for such a response both reflect the intersubjective norms which make up the tacit nature of communication and mutual compliance/consensus-seeking. Coming home late may be 6 wrong because one should show consideration and respect for others, or because agreements are important in relationships, or being able to count on someone is important. Some normative claim rooted in interpreted needs and interests (Rehg, 1994) is made by one upon the other. It was Kohlberg's seminal effort to demonstrate how the process of individually validating norms proceeded in an age-related sequential fashion. Kohlberg's Moral Judgment Interview is an open-ended discussion of standard hypothetical dilemmas frequently encountered within moral philosophy. Responses are assigned to one of six stages. For example, in one of Kohlberg's hypothetical dilemmas, the son Joe is asked by his father to render up his hard earned paper-route money so the father can go on a fishing trip. "How contemptible!" it might be said of the father. Or on the other side, "After all the father has done for the boy it is the least the son can do!" How one comes down on the issue, and the justifications or judgments offered in defense, are all potentially scorable according to Kohlberg's elaborate coding manual (Colby & Kohlberg, 1987). This particular hypothetical dilemma pits the son's property rights (i.e., entitlement) against his duty/obligation to the father in their complementary roles as father (responsible authority) and son. A scorable conventional Stage 3 judgment would be: "The most important thing a son/father should consider is to try to understand the other, respect the other's feelings, see each other's point of view, be willing to listen to each other, or think what it is like to be a child or parent." This moral judgment best reflects the "balancing perspectives or role taking" developmental capacities that emerge around early adolescence, reflective of the justice ideal at conventional Stage 3. The insight emerges at Stage 3 that self and other can simultaneously consider each other's viewpoint. What the interviewer seeks to elicit from the research participant over the course of the approximately half to three-quarter hour interview is reasoning on both sides of the issue, reasoning regarding both entitlements and obligations. A final stage score is assigned based on the type of moral judgments generated across all scorable judgments. Little research has 7 been done examining those cases where individuals repeatedly produce material on only one side of the bipolar entitlement-obligation dimension. Among the rare exceptions, Trevethan and Walker (1989) found that psychopaths used considerably more egoistic utilitarian judgments in their discussions of real-life dilemmas than did delinquents, reflective of their focus on self-interest. More indirectly, much has been said within cultural criticism about the way the language of classic liberal rights has constituted a greater social expectation of entitlement in the West, having lost sight of the reciprocal obligatory pole (Elshtain, 1993). Overall, the work of Gilligan could be seen to consist of a discussion of the (complmentary) obligatory side of Kohlberg's conventional Stage 3, by attempting to understand Stage 3 structure as expressive of a desire to "preserve relationships." It was Gilligan's initial argument that women's "concerns with preserving relationships" had originally resulted in their being downgraded as morally immature relative to men in Kohlberg's scoring system. She argued that Kohlberg's model resulted in more women being stuck at the Stage 3 relationships perspective whereas more men went on to "higher" stages of principled moral reasoning. Following a similar line of argument, Jack (1991), of whom more will be said in what follows, also linked conventional care with the (woman's) desire to preserve relationships, which she, in turn, further relates to constructs such as "anxious attachment" and "compliant connectedness." Whether these hypothesized relations actually hold are empirical questions the current study sought to address. Picking up on Kohlberg's acknowledgement that his cognitive-developmental account may fail "to map the entire moral domain" (Kohlberg, 1986, p. 500, italics added), Gilligan edited a book entitled, Mapping the Moral Domain (1988). Kohlberg's moral judgment coding manual (Colby & Kohlberg, 1987) makes no claim to having provided a complete listing of all possible moral judgments that persons are capable of making in defense of an action. This "mapping" project was intended to link her ethic of care and Kohlberg's justice ethic, calling "attention to moral judgments that did not fit the [Kohlbergian justice] definition of 'moral' and to self-descriptions at odds with the [interpersonal] concept of self" 8 (p. xvii). The current research continues the mapping process, with a particular eye to assessing existing constructs and further developing new ones inherent within the "domain" of care - or more specifically (Gilligan's Level 2) conventional care where "the 'good' woman becomes caring by becoming selfless" (Gilligan, 1998, p. 138). Where this work diverges from Kohlberg's and Gilligan's approaches is in moving toward the examination of continuous variables. The problem with any typology or categorical variable where an individual is described as not/being of Type X , be it a developmental one or not, is that considerable variance is lost. According to Skoe (1993), for example, Level 2 global stage descriptions, of the sort quoted above, an individual would be coded at a conventional care stage by justifying not leaving a relationship, for example, so as "not to let people down," or because "it would not be the religiously sanctioned thing to do." The stage structure may well be present, but the degree to which a care focus on the other involves a repudiation of self is lost. A good deal of codable variability is especially needed to provide the necessary ceiling room for the creation of clinical scales. Secondly, an examination of the facets - or stage "elements" to use Kohlbergian language - may prove them to be empirically quite distinct, even though they share an underlying common structural background. That is, facets of the conventional care construct may prove to have very distinct sets of correlates, too easily passed over when relying on a notion of a global stage. As elements within a stage level, the multiple facets of the construct of care must, it is argued, also be examined in relation to associated constructs such as dependency and attachment. G i l l i g a n a n d the V o i c e o f C a r e Gilligan's (1982) proposal of the existence of two fundamentally distinct and gender-related moral orientations, care and justice, sparked a large controversy that has not yet abated. Gilligan's move was to link a (personalist) conception of morality - a conception that has historically long stood in opposition to Kohlberg's (impartialist) tradition - with both gender and conceptions of identity associated with care and justice. 9 Gilligan argued that the masculine sense of self is bounded and separable, not containing the other as part of its identity; whereas the feminine sense of self is bound up with the other. Women, on this account, define themselves in relation to others, and men do not. Gilligan articulates the Kohlbergian justice/rights orientation in terms of separateness and the care/response orientation in terms of connectedness. "Responsibility," according to Gilligan, is understood in terms of responsiveness to the needs of the other, rather than obligation/duty based on deduction from abstract principles. The separate-self judges according to reciprocity, rules, and roles. In much of the same voice, Haan (1986) caricatures the Kohlbergian moral agent as consulting the stage structure of their development to locate the relevant general rule or principle that will decide the issue by transcending its content and its social-personal context.... When the additional judgment of self-responsibility is made, people will act in accordance with their stage.... In this view actors have a certain detachment from moral conflict, (p. 1272) Gilligan's etiological account of the gender differences is based on a same-sex parental identification process more familiar within certain psychodynamic accounts of development (Chodorow, 1978). She argues that boys are required to separate from the mother in order to identify with their father, whereas the developmental experience of girls is different, in that gender identification does not require separation from the mother. On this classical account, two distinct experiences arise for the boy and the girl. The boy, once having separated, comes to value autonomy and objectivity, whereas the girl comes to value connectedness and caring relationships. As young girls reach adolescence, however, they come to a fork in their pathway; a point in development while pressing girls to separate also heightens their sensitivity to relationships and disconnection. As Gilligan argues, Girls, because of their more acute personal encounter with disconnection at adolescence [become] alerted to problems of connection at a time in history when relationships in general have become corrupt, (p. 11) Based on a more qualitative or collaborative "listening" to 100 adolescent girls from a boarding school in the Midwest, Gilligan concludes that, for girls, adolescence constitutes a 10 special "impasse in female development" (1990, p. 9). I felt at times that I was entering an underground world, that I was led in by girls to caverns of knowledge.... What I heard was at once familiar and surprising: girls' knowledge of the human social world, a knowledge gleaned by seeing and listening, by piecing together thoughts and feelings, compelling in its explanatory power and often intricate in its psychological logic, (p. 14) On the basis of such evidence, Gilligan reports that "much of what psychologists know about relationships is also known by adolescent girls" (p. 24). Girls at this point are taught to give up their authentic voice (what they really know regarding their valuation of connection) in order to fit into a world where they are required to be "nice," "good," "caring," "helpful," "undemanding," and "unselfish." In response to such pressures, girls come to answer with the sign of repression, "I don't know," in response to questions that were once spoken with self-confidence. Consequently, the clinical interviewing approach, according to Gilligan, is necessary in order to move through the introjected voices of conventional morality. I also begin to listen with girls to the voices which they are taking in. Opening their ears to the world, listening in, eavesdropping on the daily conversations, girls take in voices which silence their relational knowledge.... Voices which intentionally or unintentionally interfere with girls' knowing, or encourage girls to silence themselves, keep girls from picking up or bringing out into the open a series of relational violations which they are acutely keyed into, such as not being listened to, being ignored, being left out, being insulted, being criticized. (1991, p. 19) The voices which precipitate this process of dissociative self-silencing involve the (patriarchal) sanction against selfishness, where "goodness" is understood as self-sacrificial care. The wall that keeps memory [read: one's authentic voice] from seeping through these covers may be the wall with the sign which labels body, feelings, relationships, knowing, voice and desire as bad. (Gilligan, 1991, p. 23) Gilligan's beliefs regarding female adolescent development directly build on her earlier formulations regarding the centrality of women's interdependent sense of self. Gilligan's In a Different Voice (1982) has now become canonical in the study of gender differences; introducing the notion of "voice" to contemporary feminist scholarship (Davis, 1994). "Voice" came to express a feminine morality, since it both referred to an ethic of care 11 and was expressed to a greater extent by women. Gilligan's work became a rallying point for feminist scholarship, transforming the very qualities historically used to declare women as inferior into virtue. Empirical Investigations Do women experience a sense of self distinct to that of men, one less bounded and more relationally defined? Several studies have sought to examine the questions raised by Gilligan and others. In what follows, empirical studies relating to the two parts of Gilligan's argument will be examined: (a) the existence of an interdependent self; and the (b) the presence of an unique "voice" or ethic of care. Women's Interdependent Self A number of researchers have pursued the question of gender differences in self-construal. McGuire and McGuire (1982), using an open-ended format with school children aged 7 to 17, found that more girls than boys expressed self-conceptions that were fundamentally social. Girls described themselves in terms of other people 50% more often, and including more spontaneous references to significant others, whereas boys included more spontaneous references to people in general. Similarly, using an autobiographical method, Clancy and Dollinger (1993) found women included within their photo albums more pictures of themselves with others and more pictures including family members. By contrast, men tended to have more pictures of themselves alone than did women. In descriptions of their ideal and undesired selves, women were also found to be more likely than men to include relationships (Bybee, Glick, & Zigler, 1990; Ogilvie & Clark, 1992). In a related vein, when participants were asked to respond to experimenter-selected attributes, men were more likely to evaluate themselves positively on those representative of independence (e.g., power and self-sufficiency), whereas women regarded themselves positively on more interdependent attributes (e.g., likability or sociability; for reviews see Maccoby & Jacklin, 1987; Simmons, 1987). In addition to gender differences found in the content of self-descriptions, using both 12 spontaneous self-descriptions and fixed questionnaire formats, men and women are also reported to differ in the centrality or importance of certain values. In a study of adolescents' self-concepts, for example, girls tended to regard interpersonal harmony and sensitivity as more important than did boys. In contrast, boys in this study regarded social dominance and toughness as more important than did girls (Rosenberg, 1986; see also Eccles, Wigfield, Flanagan, Miller, Reuman, & Yee, 1989). Similarly, in a study examining the relation of adult roles to identity, women ranked relational aspects (e.g, spouse, friend, son or daughter) as more central to their identity than did men (Thoits, 1992). Gender differences also have been found by those using various information-processing approaches of the so-called "self-referential effect." The assumption of this line of research is that individuals have better memory recall for words encoded with respect to the self. As such, words that are more readily recalled are assumed to have been more "deeply processed" (Greenwald & Pratkanis, 1984). Following these dictates, Josephs, Markus, and Tafarodi (1992) studied the hypothesis that men's and women's self-esteem is grounded in different sources. Based on the assumption that self-esteem is derived from succeeding in what is valued within a particular socio-cultural niche, they hypothesized that women with high self-esteem would recall a significantly greater number of words that involved associations with a close friend. In general, their results are said to support the view that women [with high self-esteem] have highly elaborated structures of knowledge about important others and ... the information encoded with respect to these others can be used to produce a rich, highly memorable encoding of the stimulus words in these conditions, (p. 396) By extension, one would think that the memory of women for people and relational events should be more accurate than that of men, who are assumed to have a more independent self-construal. Consistent with such expectations, a meta-analytic review conducted by Hall (1984) found that women had better recall for faces than did men, d = .34. This same gender difference in face recognition has also been found in children as young as 13 4 years old (Feldstein, 1976). In comparison with men, women recall more details of persons casually encountered in the street (Yarmey, 1993), remember more high school classmates' names and faces in the years following (Bahrick, Bahrick, & Wittinger, 1973), and recall more vivid and detailed accounts of relational events (e.g., a vacation, an argument, or first date; Ross & Holmberg, 1992). Similarly, female counsellors are reported to recall more details of their clients than do male counsellors (Buczak, 1981). By contrast, it was found that men have a greater recall of historical events (Storandt, Grant, & Gordon, 1978). Similarly, men are more likely to over-estimate the degree to which they consider their own characteristics or attributes as unique and unshared by others, a phenomenon referred to as the "false uniqueness bias" (Goethals, Messick, & Allison, 1991). Women also reported that not being forgiven by a friend would have a greater impact on their self-esteem than did men (Hodgins, Liebeskind, & Schwartz, 1996). The "Voices" of Care versus Justice Direct empirical investigations of Gilligan's claims using semi-structured interviews of real-life and hypothetical dilemmas (e.g., "Should Heinz steal the drug that would save his wife?") have not only partly confirmed her hypotheses, but have opened up new avenues for their interpretation (Walker, 1991). On the strong form of Gilligan's account - that care and justice orientations divide along strict gender lines - studies examining intraindividual consistency have found that the number of both women and men who used the same orientation across two real-life dilemmas did not differ from chance (Pratt, Golding, Hunter, & Sampson, 1988). Such findings do some damage to Gilligan's claim that "most people ... focus on one orientation and minimally represent the other" (1986, p. 10). Others report, however, that women produce a proportionately greater number of care responses on real-life dilemmas than do men (Walker, 1989; Langdale, 1986). When the type of dilemma (personal versus impersonal) is controlled, however, gender differences lose their significance (Walker, 1995). Personal dilemmas (i.e, those involving conflicts among persons with an ongoing relationship) elicit more care reasoning for both men and women. 14 Impersonal dilemmas (i.e., those involving conflicts with strangers or institutions) were found to similarly elicit more justice reasoning. What began as a possibly strict division between care and justice types of moral reasoning now appears to turn on the kind of dilemma the participant generates. The further question is then begged: Why are more women raising personal dilemmas? Another possible contributing factor to possible gender differences in rights/care orientations concerns the potentially polarizing effect inherent in the parenting role. Walker (1989; Walker, de Vries, & Trevethan, 1987) found gender differences in moral orientation among parents, but not children. Pratt et al. (1988) compared two samples of same-aged adults, half of whom were parents and half, not. This pattern was only evidenced among the parents, indicating "that the sex difference in moral orientations is of rather limited generality" (Walker, 1995, p. 97). Various meta-analytic investigations of the moral development litersture have suggested that gender differences are rapidly disappearing, supporting the socially constructed nature of gender stereotypes. Knight and colleagues (Knight, Fabes, & Higgins, 1996), however, question to what extent these causal inferences may not be an artifactual result of changes in research methodology. Based on their own meta-analyses of aggression - one of the most stereotypic of gender differences in favor of men - they conclude, "gender differences in aggression appear to be remarkably stable when changes in study characteristics over time are controlled" (p. 410). In general, Gilligan's critics have charged her theory as deficient in that it is reifying past gender stereotypes, and fails to give sufficient weight to related matters of ethnic and socioeconomic differences. In other words, Gilligan is said to have confused prescriptivity for descriptivity (Bebeau & Brabeck, 1989; Okin, 1989; Puka, 1991). It has been argued that a great deal of what has been said about women is equally true of poor people, who place less emphasis on work and self-advancement, are deferential to .those occupying higher positions of status, and appear to others as more sensitive and intuitive than rational. In 15 Pollitt's (1992) critique of Gilligan, the moral ideal of care seems to function more as a rhetorical means of advancing women's status by portraying them as morally superior. Miller (1986) has argued that because of the relative powerlessness of women in our culture, despite gains made by the Women's Movement, women must be sensitive and responsive to others, especially others who have power over them. Persons in positions of power may promote rights and rationality, whereas those in subordinate positions will use more indirect means of attaining ends by advocating connection and concern (Hare-Mustin & Marecek, 1988). In summary, recent investigations of Gilligan's understanding of women's interdependent self-construal and moral orientation of care, using open-ended questions, have been partly supportive; women do define themselves more interpersonally and produce more care reasoning when discussing real-life dilemmas than do men. By contrast, other research has highlighted a number of possible confounds: being in the role of a parent may have a polarizing effect on the type of orientation expressed, and particular types of dilemma (e.g., personal versus impersonal) may "pull" (Krebs, Vermeulen, Carpendale, & Denton, 1991) for one orientation over another. Studies using standardized questionnaires have generally failed to find gender differences (Friedman, Robinson, & Friedman, 1987; Walker, 1995). Gilligan's (Brown, Debold, Tappan, & Gilligan, 1991; Brown & Gilligan, 1992) more recent move to a "narrative" or qualitative approach to investigating the voice of care, has been found "vague and unreliable" by some (Wark & Krebs, 1996). The problem of successfully operationalizing the "care" construct has been an ongoing one. Efforts, like those of Jack (1991), to develop an objective measure of Gilligan's care construct are needed if the many confounds which invade the empirical investigation of gender differences are to be better understood. Silencing the Self With Jack's book, Silencing the Self (1991), and the introduction of a paper-and-pencil measure of self-silencing (The Silencing the Self Scale) or what has here been called 16 conventional care, Gilligan's ideas have been used to explore a variety of clinical problems ranging from depression to drug abuse. At present some 10 articles, 26 conference papers, and almost 50 theses and dissertations have included Jack's measure (D. Jack, personal communication, April 1998). At her Kohlberg Memorial Lecture Gilligan stated, In Silencing the Self: Depression and Women (1991), Dana Jack brought the process of dissociation into the centre of this conversation by showing how depressed women silence themselves.... we realise that we can not know what we know, not feel what we feel, not say what we mean, nor care about what we care most deeply about. (1998, p. 132) Elsewhere, and in a special feature responding to her critics, Gilligan cites Dana Jack's study of depression and women (Silencing the Self 1991) and the scale derived from that study... [as being] among the first of a growing number of innovative, empirical studies in developmental and clinical psychology that have been stimulated by my work on voice. (1994, p. 421) Jack's Silencing the Self Scale, however, has undergone very little in the way of construct validation and is typical of scales in the field. A number of points will require clarification before this burgeoning population of studies approaches intelligibility. The first point involves a question about the sorts of theoretical interpretation that are made of the self-silencing "data." On the one hand the Silencing the Self Scale (STSS) is said to "not overtly refer to the feminine role but only to its imperative" (Jack & Di l l , 1992, p. 98); yet, on the other hand, it is described as "an instrument to investigate gender-specific schemas ... about how to create and maintain safe, intimate relationships [which] lead women to silence certain feelings, thoughts, and actions" (pp. 97, 98; italics added). Jack argues that the centrality of relationships for women's self-identity, in conjunction with gender role norms urging women to be compliant and unselfish, places women at unique psychological risk. While the results across several studies using mixed samples of both men and women report gender differences in favor of men doing the greater proportion of self-silencing (e.g., Cowan, Bommersbach, & Curtis, 1995; Jack & Di l l , 1992; Thompson, 1995), discussion continues to revolve around engendered self-in-relation theory: "High scorers [on the Silencing the Self Scale] ... reflect greater pressure to fulfill norms of the 'good woman'" 17 (Jack & Di l l , 1992, p. 99). Are men who self-silence to be understood as fulfilling norms of the "good woman"? Interpretation of various findings of high levels of self-silencing has turned on the conditional adverbial clause, "When I am with my friends," that appears in many of the stems within the STSS. The assumption, it would appear, is that these behaviors are for the purpose of maintaining already existing relationships, albeit ones in which individuals are not very contented. Jack speaks of "compliant connectedness" (1991, p. 40) in a way that resembles anxious attachment. Depressed women's statements such as, "I have learned, don't rock the boat with my partner" ... show their conscious awareness of making themselves appear passive or compliant for an intended effect: to keep outer harmony, to preserve relationship. (Jack, in press, p. 9) The attachment literature, however, regards behaviors involving compulsive caretaking and anxious attachment as focused primarily upon a particular close attachment figure. Would the above description not also fit for the dependent woman of an abusive partner with the attendant motivation not to seek intimacy, but simply to maintain harmony for the preservation of life and limb? Attachment and anxious attachment, in particular, are also different constructs reflecting some general strong inclinations towards "affiliation." There seems to be some potential for confusion between the supposedly distinct domains of self-silencing and "anxious attachment," or other more general forms of social anxiety (Bartholomew & Battel, 1998). In a to-be-published chapter (Jack, in press), much of the discussion of the role of self-silencing revolves around women's vulnerability within our culture and the relational-self. On the few brief occasions when men's self-silencing is spoken of, it is in the context of a yet broader domain - that of work. Dan's [one of the interviewees] litany of shoulds contain not only an implied perfectionism, but also an image of selflessness focused on [his] professional role: he should be "selflessly" married to medicine, (p. 29) Self-silencing seems here to be no longer for the purpose of maintaining relationships but 18 occurs in the more impersonal context of individuals at work. Dan says, "I feel that I need to put on my professional face and my caring face" when he is with his patients (p. 29). Jack is surely correct when she writes, "researchers must look behind self-silencing for its gendered meanings and its relational intent" (p. 9). Regarding men's greater tendency of "withdrawal through silence or passive resistance" (Gottman, 1994), Jack's extracted intent is, "when men self-silence, they may intend to create distance and control interactions in relationships" (p. 9). If men's "stonewalling" is also a form of self-silencing then this represents a very different form of behavior from how self-silencing is described elsewhere; this represents a cold-avoidant form of interpersonal behavior. It may be that self-silencing reflects a great many interpersonal behaviors beyond anxious attachment and general social anxiety, but self-in-relation theory provides no suitable non-gender-specific interpretation. Part of the confusion enters when the designator "self-silencing" is intended to reflect both the global construct indicated by the title of the scale, Silencing the Self, as well as one of the subscales, Silence. Should self-silencing (i.e., conventional care) prove to be a multifaceted construct, this only adds to difficulties in gaining interpretive clarity. A second point of concern has to do with the dangers of any theoretical/empirical enterprise becoming too insular and self-referential. If men also self-silence, then a broader conceptual framework will be needed to give some coherent account of this fact. Generally, what is needed is to locate the Silencing the Self Scale, and other scales of this sort, not only in conversation with one another, but more importantly within a nomological net of broader constructs whereby the instrument could either be verified as an independent contributor, or found redundant with existing variables. It is here suggested that the dependency and attachment literatures form just such a starting point for such an investigation. From Table 1 it can be seen that the thematic content of items within the various STSS subscales (in the left column) show considerable resemblance with those of other dependency/care/attachment measures (in the right column) such as the Sociotropy Autonomy Scale - Pleasing Others Subscale (SAS-PO; Beck, Epstein, Harrison, & Emery, 1983), the I] 19 Table 1 Comparison between Self-silencing Subscales and Measures of Care and Dependency Self-silencing SAS-PO • When friends' opinions conflict with • People can pretty easily change me mine, rather than asserting my own point even though I thought that my mind was of view I usually end up agreeing with already made up on a subject. them. IBT • I find it hard to go against what others think. Care UCS • In my friendships my responsibility is • For me to be happy, I need others to be to make the other person happy. happy. External SPS-SD • I often feel responsible for my friends' • I consider myself a failure if I can't act feelings. the way others want me to. Unmitigated Communion Scale (UCS; Helgeson & Fritz, 1998), the Social Perfectionism Scale - Self-directed (SPS-SD; Wiebe & McCabe, 1998), and the Irrational Beliefs Test -Demand for Approval Subscale (IBT; Jones, 1969). What is the relationship between self-silencing (i.e., conventional care) and the other dependency/attachment constructs? There would appear to be a constellation of constructs here that invite comparison and contrast. Test and theory developers must be continually open to the possibility that the operationalized ground that their constructs are intended to furrow may be narrated from very different theoretical viewpoints. A structural clarification is needed to examine both the consistency within the proposed constructs, as well as their interrelations within a larger nomological net. For example, an item used in the current study reads "I am afraid of making mistakes in conversations" (Self-directed Social Perfectionism Scale). While the item was intended by authors Wiebe and McCabe (1998) to be a motivational construct aimed at seeking social perfection, it could just as easily be associated with avoiding relationships, or a dependency construct such as Fear of Negative Evaluation 20 (e.g., "I feel very upset when I commit some social error"; Watson & Clark, 1984). The relationships among the constructs of care, dependency, and attachment will require considerable clarification. Dependency, Attachment, and Care The constructs of care, dependency, and attachment all have been variously used with overlapping reference. Thompson states that Gilligan's notion of compliant connectedness "resembles anxious attachment, [and] is characterized by compulsive caretaking, pleasing the other, and inhibition in self-expression" (Thompson, 1995, p. 338). From Bowlby's perspective, however, the construct of compulsive caretaking was seen to be distinct from anxious attachment, and both attachment constructs were thought to be markedly distinct from dependency (Bowlby, 1973/1991, 1980/1991). What care, dependency, and attachment constructs do share, however, is a conception of persons as fundamentally seeking attachment objects and relations, as opposed to attachment as a drive derivative (Freud, 1930/1955), or as a generalized association with sex and/or food (e.g., behaviorism). The construct of dependency has its roots in the psychoanalytic image of the infant-caregiver relationship. Freud believed that, an excess of parental affection does harm by causing precocious sexual maturity and also because, by spoiling the child, it makes him [sic] incapable in later life of temporarily doing without love or of being content with a smaller amount of it. (1894/1955, p. 223) In this light dependent persons are characterized as immature and childlike. Atop this basic metaphorical picture of the child's material dependency needs, when referred to adults the importance of identifying additional component constructs has been frequently recognized. The range and type of components or facets used to define the whole, however, varies. Zuckerman, Levitt, and Lubin (1961) defined dependency as involving succorance (attention and approval seeking), deference (a tendency to subordinate oneself to others and inhibit self-assertion), and abasement (involving self-blame and guilt). The Navran (MMPI) Dependency Scale possesses largely the self-critical facet of the construct 21 (Birtchnell, 1984). The list of constructs that has come to be associated with dependency is a lengthy and overlapping one: suggestibility, interpersonal compliance, conflict-avoidance, pessimism, self-doubt, deference, abasement, emotional reliance upon others, lack of social self-confidence, conformity, help seeking, need for approval, need for guidance, passivity, and pleasing others. Dependent persons are said to be interested primarily in avoiding responsibility and interpersonal conflict; they try to compensate for their sense of helplessness by obtaining support of others through occupying a lower status or follower position. (Assor, Aronoff, & Messe, 1981, p. 790) Against this largely disparaging psychological backdrop it was important for Bowlby (1969/1991) to make a distinction between dependency and attachment. Logically, the word "dependence" refers to the extent to which one individual relies on another for his [sic] existence, and so has a functional reference; whereas attachment as used here refers to a form of behavior.... whereas dependency is maximum at birth and diminishes more or less steadily until maturity is reached, attachment is altogether absent at birth and is not strongly in evidence until after an infant is past six months. The words are far from synonymous, (p. 228, italics added) Bowlby sought to reserve the concept of attachment for behaviors directed towards one or a few particular attachment figures. The early parent-child attachment forms the blueprint for subsequent relationships, guiding behaviors and expectations in later romantic relationships in particular. Similar to children, securely attached adults will seek contact with a significant other in the face of stress, be comforted by the presence of the attachment figure, and experience distress at the threat of losing the attachment figure. On this account, the role of an attuned parental figure needed to be distinguished from the notion of "spoiling," which is associated with dependency. It may, however, be questioned whether Bowlby's definition of dependency, as involving material "existence" (see above italics), isn't excessively narrow (Birtchnell, 1984). Bowlby (1969/1991) proposed that dependency should be regarded as a fusion of Murray's (1938) need for affiliation ("to form friendships and associations," p. 83) and need for succorance ("to seek aid, protection, and sympathy," p. 83). Bowlby associated 22 Figure 1 Dependency as an Aspect of Attachment-detachment and Directiveness-receptiveness Dimensions Note. From "Attachment-detachment, Directiveness-receptiveness: A System for Classifying Interpersonal Attitudes and Behaviour," by J. Birtchnell, 1987, British Journal of Medical Psychology, 60, p. 18. Copyright 1987 by the British Psychological Society. overdependency with insecure, more specifically anxious, forms of attachment. Most relevant to the current research, Birtchnell (1987) proposed understanding dependency as a broad multifaceted construct locatable within two fundamental dimensions of human functioning: directiveness versus receptiveness and attachment versus detachment (see Figure 1). As such, Birtchnell broadly locates dependency between "the tendency to stay close" and an "inclination to assume a submissive attitude" (p. 18). Drawing from attachment theory, Birtchnell defines attachment as "a need for proximity and a fear of being alone," directiveness as "an inclination for giving or doing things to others ... characterized by a fear of being controlled or taken over," and receptiveness as "an inclination to assume a recipient attitude towards others and is characterized by a fear of assuming responsibility" (p. DIRECTIVENESS (inclination to assume a dominant attitude) DETACHMENT (tendency to remain distant) ATTACHMENT (tendency to stay close) DEPENDENCY RECEPTIVENESS (tendency to assume a submissive attitude) .25). 23 One of the problems with the above proposal is that, in mapping out the domain of dependency as involving an entire quadrant, and some area in the quadrant immediately above (as will later be discussed), Birtchnell contains within the one construct, variables which are orthogonal (i.e., attachment and directiveness are orthogonal dimensions). One solution, and the one taken here, is to propose a division of constructs based on their position within further divisions of the quadrant. More will be said of this proposal below. Motivated in a similar way to Bowlby, Gilligan sought to distance the ethic of care and women's associated interdependent sense of self from Freud's conceptions of morality and dependence. Because they were judged to be destitute as a result of an incomplete resolution of the Oedipus complex and thus not having a properly developed superego, Freud inferred that women "show less sense of justice than men, that they are less ready to submit to the great exigencies of life, [and] that they are more often influenced in their judgements by feelings of affection or hostility" (1925, pp. 257-258). It was through Chodorow's (1978) re-interpretation that aspects of Freud's dependency and moral inferiority became the virtues of care: "Girls emerge with a stronger basis for experiencing another's needs or feelings as one's own" (Chodorow, 1978, p. 167). In light of what has been said above, what is to be made of the rapidly expanding data using Jack's Silencing the Self Scale and scales of its kind? Critics of Gilligan have charged her with being overly schematic. Tavris (1994) is doubtful that: girls and women speak consistently across situations, that adolescent patterns should determine later adult outcome, that self-silencing is unique to or common among women, that "power" is male and "powerlessness" female, and that self-silencing is strictly a bad thing. Are men who self-silence to be understood as fulfilling norms of the "good woman" or should the conventional sanction against selfishness be moored elsewhere? Questions multiply, with an absence in the field of sound measures by which to undertake their investigation. It can be anticipated that dependency, attachment, and (conventional) care are locatable within the same semantic space. If these constructs and their respective scales are 24 not located within some broader nomological net and the offerings of Gilligan's self-in-relation theory not interpreted within a larger psychological framework, only further polarization and a lack of clarity will likely result. Problems arise when any interpretive community becomes too insular. Contratto comments, "her [Gilligan's] writing and that of the other members of the Harvard Project has become increasingly self-referential" (1994, p. 368). The current study sought a structural clarification across a superordinate set of terms, or higher point of view, by which dependency, care, and attachment might be discussed simultaneously. Future Directions and Finding Common Ground One of the primary correctives that Gilligan introduced to moral psychology was to stress the importance and complexity of the interpersonally embedded particular, as captured in the metaphor of "voice" - over against the tendency to moral abstractions and universal generalizations. Favoring localizations of "voice" over the universal and general, however, risks losing sight of possible common grounds for discussion. There has been considerable consensus regarding the broader framework from within which the constructs of attachment, dependency, and care should be studied. Beck (1983) speaks of sociotropy versus autonomy, Blatt (1990) interpersonal relatedness versus self-definition, and Birtchnell (1987) attachment versus directiveness. Earlier Bakan (1966) had proposed the existence of two independent modalities, agency and communion. Agency refers to the character of being a differentiated individual, and is expressed in strivings for mastery and power. Communion refers to the character of being connected with a larger social or spiritual whole, and is expressed in strivings for intimacy and connectedness with the larger whole. Similar to Birtchnell's figure (see Figure 1 above), Figure 2 presents a schematic view of the two dimensions of agency (A+) and communion (C+) and their conceptual opposite poles, passivity (A-) and disconnection (C-). These two, agency and communion, can be found to represent fundamental dimensions across a variety of worldviews, languages, and psychological theorizations (Wiggins, 1991). 25 Figure 2 Structural Representation of Agency and Communion II. hostile dominant (C-) DISCONNECTION (hostility, disaffiliation) III. hostile submissive (A+) AGENCY (mastery, power) I. friendly dominant (C+) COMMUNION (intimacy, union) IV. friendly submissive (A-) PASSIVITY (vulnerable, submission) Beyond the speculative figures of Birtchnell and Bakan, Timothy Leary (1957) was one of the first to actually operationalize the "interpersonal circle" - which would later go through a number of subsequent revisions (Benjamin, 1974; Carson, 1969; Kiesler, 1983). There is considerable agreement among interpersonal theorists that the best structural model for representing personality dispositions and interactions is a two-dimensional circumplex in which variables are distributed continuously around the orthogonal dimensions relabelled dominance versus submission (DOM) in place of AGENCY, and nurturance versus hostility (NUR) in place of COMMUNION (Carson, 1969; Kiesler, 1982; Leary, 1957; Sullivan, 1953). Employing both rational and empirical approaches, Wiggins (1979) classified 817 of Goldberg's (1978) 1710 adjective traits into 16 interpersonal clusters theoretically suggested by the circumplex model. Interpersonal adjectives were those determined to describe: "dyadic interactions that have relatively clear-cut social (status) and emotional (love) 26 consequences for both participants (self and other)" (p. 398). The Wiggins model has been validated and employed across numerous studies (Pincus & Gurtman, 1995). Although several other broad structural models of personality have been proposed (e.g., Eysenck, 1994; Tellegen, 1985), the interpersonal focus of the current study fits well with the Interpersonal Circumplex (Leary, 1957; Wiggins, 1979), a model which has additionally shown convergence with the Five-Factor Model (McCrae, 1989). Wiggins and his colleagues, as well as McCrae and Costa (1989), have demonstrated that two of the more robust factors of the Five-Factor Model, Extraversion and Agreeableness, but for a modest transposition of axes, are isomorphic with the Dominance and Nurturance dimensions, respectively. What divides the two most broadly employed models, the Five-Factor Model and the Interpersonal Circumplex, is their understanding of the additional three factors (i.e., Openness, Conscientiousness, and Neuroticism) regarded as primary to the former. In the Five-Factor Model they represent three additional factors, equal in their own right as factors alongside the other two. From the perspective of the Interpersonal Circumplex, these additional three factors are considered as imparting an additional "coloration" or character upon the more central circumplex "dimensions" of dominance and nurturance (Trapnell & Wiggins, 1990). However the conceptual relations among the five super-factors are understood, their uncovering represents significant conceptual and empirical progress in the field of personality psychology. Whereas the 1980s saw a tremendous disavowal of even appearing to be a "trait theorist" - regarded "like witches of 300 years ago" as Jackson and Paunonen (1980, p. 523) derisively observed - recent convergences are said to now provide "a theoretical structure of surprising generality" and "a good answer to the question of personality structure" (Digman, 1990, pp. 418, 436). The five factors seem to cover a vast conceptual space, integrating Cattell's 16 factors, Eysenck's "big three," Murray's 20 needs, Guilford's temperaments, Jung's types, and Block's California Q-set (McCrae, 1989). Not only do these five factors provide a compelling framework for building personality measures 27 that are capable of representing the domain of personality broadly and systematically, but such a framework enables researchers to locate a manifold of constructs and measures within a meaningful conceptual space thereby enabling their comparison and contrast, and bringing clarification to what Adelson (1969) referred to as "a disconcerting sprawl." The possession of such a model will also provide a good starting point for investigations into the biological substrate of human universals (Livesley & Jang, 1993). Numerous cross-cultural studies have to varying degrees largely supported the Five-Factor Model in at least six different languages (English, German, Japanese, Chinese, Tagalog [Filipino], and modern Hebrew). Goldberg writes: They [the five factors] suggest that those who have contributed to the English lexicon as it has evolved over time wished to know the answer to at least five types of questions about a stranger they were soon to meet: (1) Is X active and dominant or passive and submissive (Can I bully X or will X try to bully me)? (2) Is X agreeable (warm and pleasant) or disagreeable (cold and distant)? (3) Can I count on X (Is X responsible and conscientious or undependable and negligent)? (4) Is X crazy (unpredictable) or sane (stable)? (5) Is X smart or dumb (How easy will it be for me to teach X)? Are these universal questions? (1981, p. 161) Within a hierarchical arrangement the interpersonal circle may be conceived as integrating a series of circularly arranged subdimensions within a superordinate factor or interpersonal DOM/NTJR space. The number of subdimensions - reflected in the angles of separation between the various vectors - are understood as representing meaningful differences in interpersonal behavior (e.g., Strong, Hills, Kilmartin, DeVries, Lanier, Nelson, Strickland, & Meyer, 1988). While some investigators have found even small angles of separation to be meaningful (Benjamin, 1974; Kiesler, 1996), most investigations have opted for a circumplex division into octants (i.e., eighths). Figure 3 shows how various interpersonal traits are distributed around the circumplex axes of dominance (DOM) and nurturance (NUR) in Wiggins' (1995) Interpersonal Adjectives Scales; angular locations for each are given with the right pole of the x-axis (NUR) representing 0°. The circumplex approach describes healthy interpersonal functioning, including the ability to care, as marked by the flexible enactment of any of the trait descriptors when the 28 Figure 3 Items and Circular Locations for the Interpersonal Adjectives Scales (DOM) Assured-Dominant (PA) 95* Firm (.269) 107* Dominant (.423) 108* Forceful (.377) 114' Domineering (.342) Arrogant-Calculating (BC) 130* Cocky (.254) 136* Crafty (.393) 137* Cunning (.456) 140* Boastful (.200) 142* WDy (.385) 143" Calculating (.205) 144' Tricky (.427) 146' Sly (.452) Cold-Hearted (OE) 159' Ruthless (.291) 175' Ironhearted (.327) 179' Hardhearted (.364) 179* Uncharitable (.286) 180* Cruel (.292) ) 186' Cold-hearted (.358) 187* Unsympathetic (.369) 196' Warmthless (.373) Aloof-Introverted (FG) 215* Uncheery (.358) 215" Unneighborly (.336) 217* Distant (.337) 218* Dissocial (.429) 2 2 1 ' Unsociable (.452) 222* Antisocial (.389) 229* Unspariding (.352) 242' Introverted (.406) 89* Persistent (.171) 87* Assertive (.443) 84 ' Self-confident (.398) 8 1 ' Self-assured (.391) (NO) Gregarious-Extraverted 66' Extraverted (.460) 57' Outgoing (.570) 50" Enthusiastic (.357) 45' Perky (.328) 40' Jovial (.303) 34* Neighborly (.308) 33 ' FriefxSy(.471) 30' Cheerful (.414) (LM) Warm-Agreeable 9* Kind (.229) 6* Sympathetic (.310) 6* Tender (.368) 4* Charitable (.290) 360' Tenderhearted (.376) 360' Gendehearted (.390) 356" Accommodating (.182) 349' Soft-hearted (.337) (JK) Unassuming-Ingenuous 321 ' Unsly(.319) 318* Uncunning (.348) 316' Unwily(.279) 313* BoasUess(.171) 311' Uncrafty (.306) 305' Uncalculating (.213) 295' Undemanding (.200) 294* Unargumentative (.198) (NUR) (HI) Unassured-Submissive 255' Timid (.330) 283' Unaggressive (.399) 256' Bashful (.305) 279" Untold (.314) 258* Shy (.373) 276' Unauthoritative (.321) 260' Meek (.306) 272' Forceless (.314) Note. From Interpersonal Adjectives Scales: Professional Manual, by J. Wiggins, 1995, p. 23. Copyright 1995 by Psychological Assessment Resources. context deems them appropriate; neither chaotic nor rigid. Well-being is characterized by a moderate amount or "intensity" of any given behavior. Intensity is operationalized as the distance (i.e., vector length) from the center of the circumplex. Lorna Benjamin adds a 29 qualitative dimension in preferring the Western cultural ideal of the friendly-dominant quadrant (upper right): "behaviors that are friendly and moderately enmeshed or moderately differentiated are normal.... and [Quadrant I] represents an optimal position in relation to others" (1996, p. 253; see Figure 2). While hostile behaviors can be normal in particular contexts, normality in the ideal or prescriptive sense would be located in the Quadrant I. So Bakan (1966) believed that if gone unmitigated by a positive sense of communality, a strong sense of agency would be harmful to the individual and society. Similarly, the desirable aspects of agency must mitigate against an overpowering sense of communality if the individual is to function successfully. Within the interpersonal circumplex, the first quadrant (0°-90°) represents the friendly-dominant region and is comprised of a "blend" of agency and communion. Prescriptivity is never value-free, nor need "adaptiveness" be taken as the strict ideal. Empirical support for the hypothesis that flexible responding is highly functional/ healthy (i.e., behavioral flexibility) is far from conclusive (Paulhus & Martin, 1988). In addition to accommodating both the descriptive and morally prescriptive use of the circumplex, as a superordinate semantic viewpoint it also has the potential to synthesize previously conflicting constructs. Gilligan has been criticized for contrasting connectedness and autonomy: "the message that readers often take from this view is that autonomy has no meaning for women - is somehow beneath them, beyond them, or unnatural to them" (Berlin & Johnson, 1989, p. 79). Rather than viewing autonomy and connectedness as bipolar dimensions marking men and women, respectively, Berlin and Johnson argue that autonomy is necessarily a part of mature relatedness, that it is precisely the mix of autonomy and warm bondedness that transforms resentful self-sacrifice and submission into generous, attuned, and mutual interaction, (p. 79; italics in original) These authors also consider the first quadrant as representing ideal healthy functioning. Gilligan has written about the importance of women bringing their needs and abilities into relationships but has avoided the word autonomy which she tends to equate with disconnection. 30 Of the two most widely used structural models of personality, the Five-Factor Model and the Interpersonal Circumplex, the latter with its orthogonal dimensions of dominance and nurturance seems ideally suited to the detailed investigation of the conventional care constructs. The present study goes beyond circumplex analyses with the inclusion of the three additional super-factors, Conscientiousness, Neuroticism, and Openness; understanding, however, the two interpersonal dimensions as most salient. In addition to providing a parsimonious description of interpersonal behaviors and a moral prescription of what interpersonal behaviors are deemed desirable, investigators have also begun to suggest an important prescriptive role for the circumplex in test development. This study will employ the interpersonal circle in both these aspects, discussing the latter in more detail in the following chapter. Clinical Applications Gilligan's seminal efforts were intended as a means of better understanding the experiences of women, who according to Gilligan had not been adequately represented in the psychological literature; especially experiences which might provide clarity to problems which uniquely or predominantly affect women, such as abortion and eating disorders. Bruch's (1973) seminal work in the field of eating disorders conceptualizes anorexia nervosa as the "desperate struggle for a self-respecting identity" (p. 250). Sounding very much like the care construct of external self-perception (Jack, 1991), for example, Bruch describes eating disordered women as relying upon the ideas of others to explain their own behavior. Alienated from their own experience, they are constantly: "double-tracking ... pursuing one's own thought while trying simultaneously to figure out parental motives and reactions and continually monitoring others' responses" (p. 62). The dissociated voices of eating disordered women sound not dissimilar to women Gilligan reports interviewing. One of Bruch's clients said, "I really feel I am not myself and this is really sick, this not being myself, not my body or person, not really a human being" (p. 71). Bruch argued that the basic misperceptions of these women involves a fundamental 31 alienation from self and others. As one woman stated: "I have this unbelievable fear of people not liking me" together with its counterpart, "I'm not worthy of their liking me" (p. 95). One of the links among the studies of gender, identity formation, and the problems that seem to largely afflict women, is the puzzle regarding the decline of self-esteem among young girls in early adolescence. Whereas judgments of self-worth remain largely stable for boys, some studies have shown a marked and gradual decline among girls (Harter, 1992). One of the major developmental tasks of adolescence involves the emergence of a sense of personal identity, including a sense of what it is to be a false-self or "not real" (Blasi & Milton, 1991; Harter & Lee, 1987). Notions of a true inner-self emerge, including a sense that one should "be true to" one's self. Adolescents speak of being a "phoney," "not expressing their true feelings," or "playing a roler"'There is a concomitant set of insights that emerges as to the motivational components of such behavior. The first motivational cluster views false-self behavior as an attempt to gain the approval of others, impress peers, and advance one's social relations. Another motivational cluster adolescents mention involves the attribution of others not liking them, their not liking themselves, and doubting their own true self. In one study comparing adolescents who reported high levels of false-self behavior with adolescents who did not report a need to conform for the purpose of gaining acceptance, Harter (1992) found that false-self behaviors were largely self-defeating. Rather than gaining the approval of peers, adolescents with high false-self behavior reported less peer approval and greater hopelessness about being liked or ever becoming likable. False-self adolescents also reported lower self-esteem and reported not knowing their true self. Correlational data, however, do not allow us to determine whether becoming a false-self undermines one's self-worth or whether poor self-worth and self-trust motivate an individual to adopt others' perceptions as one's own. What is striking is the relationship between self-worth and evaluations of personal 32 appearance with respect to the thinness ideal - now regarded as one of the definitive "core" psychological factors in eating disorders. Correspondence between the outer physical self which others see and the inner psychological self are closely tied. In study after study, at any developmental level we have examined, including older children, adolescents, college students, and adults in the world of work and family ... we have repeatedly discovered that self-evaluations in the domain of physical appearance are inextricably linked to global self-esteem. (Harter, 1992, p. 117) Correlations across the life-span remain very strong (rs = .70 to .80); self-identity has virtually become narrowed to that shallow reflection contained in the mirror. Harter (1992) finds that while ratings of perceived appearance remain stable for males from Grades 3 to 11, females' evaluations drop steadily across the same period from initially similar levels to those of the boys in Grade 3. Self-esteem, while dropping for females across the same period, does not do so as dramatically as perceived appearance. The relationship, or more specifically, the perceived direction of the relationship between physical appearance and self-esteem may be the mediating factor. Adolescent participants were asked to decide which of two orientations was most descriptive of them: • If others like or approve of me (first), then I will like and approve of myself. • If I (first) like and approve of myself, then others will like and approve of me. (Harter, 1992, p. 123) Those who chose the first orientation as self-representative - their self-esteem was an adoption of the judgments of others, or what Susan Harter calls the "looking-glass self" (Cooley, 1902/1964) - reported lower levels of self-esteem. Those who chose the second option as most representative of themselves - self-acceptance would come from the self first -reported significantly higher levels of global self-esteem. Level of self-esteem has proven the only consistent predictor of treatment outcome for eating disorders, both at the end of treatment and at repeated follow-ups (Fairburn, Kirk, O'Connor, Anastasiades, & Cooper, 1987). The links among inner self, outer self, and peer perceptions can make relationships with peers intense. Seligman (1995/1996) provides a prototypic illustration. 33 Christine jumps to conclusions when she sees her friends passing notes. She assumes that the notes say bad things about her.... When Laurie tries to talk with her, she rejects her.... In addition she refuses to look at the notes, she remains sure that the notes were about her.... If Christine continues to accuse her friends of being cruel and then withdraws, regardless of the accuracy of her accusations, her friends will retreat.... Boys, on the other hand, often get into physical fights, (p. 233) With the cognitive-developmental emergence during early adolescence of more sophisticated role-taking tasks comes the capacity for greater moral and relational elaboration. Kohlberg's (1986) conventional moral Stage 3 emerges at this time, allowing for the simultaneous and mutual conception of entitlements and obligations within relationships. But while the emergence of subsequent stages brings with them advantages, they also bring with them certain risks as well. An increase in self-consciousness and potential false-self behavior emerges. Taking up the perspective of the "other" - and by the "other" I mean to include the broader cultural conception of one's various engendered role identities - can be very self-effacing. Where those cultural standards are unrealistic, as in the thinness ideal held out for young women, the discrepancy between the actual and the ideal will result in lowered self-esteem. Further, insofar as women can be said to have a more interdependent sense of self, a culture's lack of appreciation of those attributes will likely also be encountered as repressive. Bruch's (1973) observations of women with eating disordered behaviors reflects a very similar picture of self-consciously "double-tracking," judging oneself from the perspective of the other, and false-self behavior where one's sense of self-worth is made contingent upon the viewpoint of the other. Where self-worth is contingent upon the other, considerable energy will be spent in seeking others' approval. In Chapter 3 the specific application of these ideas to eating disorders will be picked up in more detail. Summary and Overview The construct of conventional care is best understood at the intersect of both terms, and more broadly reflective of what has been called the Kohlberg-Gilligan debate. The contextualism and emphasis on care in relationships which Gilligan stresses was already present in the wider philosophical debate raging at the time. But what distinguishes her work 34 from the moral philosophy that preceded it is her linking of the contextual (or personalist) ethic, focusing on empathy and care within relationships largely, with the psychological emphasis on an interdependent self. Her final move is in associating the interdependent self and care with gender. Evidence was reviewed supporting the gender-relational self link as well as the gender-care orientation link. Recent research, while confirming these associations, has also pushed the question back further opening up the explanation of gender-differences to other potential factors beyond a strictly developmental same-sex identification paradigm. Continuing efforts to operationalize Gilligan's constructs have taken place outside of a process of more rigorous psychometric evaluation and have been thus criticized for being insular. The empirical component of the current research seeks to locate itself within a larger framework where the superordinate dimensions of agency versus communion, directiveness versus attachment, or dominance versus nurturance, are viewed as providing fundamental anchors not only of personality, but also of human development (Blatt, 1990; Kegan, 1982). The largely independent care, attachment, and dependency literatures suggest a variety of subordinate constructs that are in need of construct examination in relation to one another, over a third superordinate set of terms which the Interpersonal Circumplex is here suggested as providing. As an overview framing the considerable detail that is to follow, largely in Chapter 2, five broad goals or tasks will be delineated. The first task then, involves an exploration of the construct of conventional care, Gilligan's Level 2, as it relates to the construct of dependency (and thus submissiveness). In that dependency has come to be viewed as multifaceted- Pincus and Gurtman's (1995) own exploration was published under the title "The Three Faces of Dependency" expressed within three octants of circular semantic space -some such similar division is anticipated for the construct of conventional care. (The term "circle" or "circular" will be used synonymously with "circumplex," referring to the semantic space created by the two axes of dominance and nurturance.) Indeed the parallel tracks of 35 conventional care and dependency are likely to cross paths at various points, with the hope of mutually informing one another. This is a move of considerable importance given the frequent critique of insularity within the care literature. The second global task of the present study involves a rather detailed exploration and assessment following the proposal of circular structure test criteria to be used in conjunction with traditional criteria in scale development. While such criteria have been employed in not dissimilar ways and even suggested to test developers, none have thus far been proposed in conjunction with traditional substantive approaches, and none prior to the current study been employed in the actual construction of test measures. By these criteria several scales, entitled the Conventional Care Scales, will be developed and assessed together with related care and dependency scales which have already been introduced. The current research involves a continuation of the mapping process, with a particular eye to assessing existing constructs and further developing new ones inherent within the domain of care - or more specifically conventional care where "the 'good' woman becomes caring by becoming selfless" (Gilligan, 1998, p. 138). In addition to extending the number of facets subordinated within the traited or self-directed dimension, two further dimensions of conventional care will be examined: other-directed and socially prescribed conventional care. These three dimensions are those frequently employed within interpersonal theory and clinical practice (Benjamin, 1974, 1996; Sullivan, 1953). What distinguishes the three dimensions are the different objects towards which expectations of care are directed: towards one's self (self-directed), towards others (other-directed), and the perception of care expectations from others (socially prescribed). Each dimension is expected to have unique correlates. Third, in addition to the examination of conventional care in relation to dependency constructs, further construct validation and exploration will be undertaken examining conventional care in relation to measures of well-being, interpersonal capacities, relationship satisfaction, and the false-self. The first three goals will be undertaken in a lengthy first 36 study reported in Chapter 2. A second study is intended to confirm the factor structure of the scales developed in Study 1. Fourth, and in line with the recent shift to the examination of real-life over abstract hypothetical dilemmas (Walker & Hennig, 1997), this dissertation seeks also to locate itself within an applied behavioral context examining the role of conventional care in predicting problems with which more women are identified, eating disorders. The role of conventional care, which has not previously been studied in the eating disordered population outside of this data set, is examined against perfectionism, for which there is a large literature demonstrating its association with eating disorders. One focal area of developmental concern involves the large number of young women whose sense of self-worth is constituted largely on the perception they have of their own external appearance, particularly weight and shape. Judgments of self-worth become the more punitive as culture itself extols an increasingly unrealistic ideal of female beauty and thinness. One aspect of this is the press towards, and the meanings that get attached to, an unrealistically thin female body. Dieting has become virtually normative in Western culture with women reporting highly negative self-views largely based on weight and shape. Identity development has long been viewed as the core personality issue involved in eating disorders, facilitating the internalization of our culture's fixation on thinness. Deeply implicated is the possession of a false-self expressive of the belief that for one to like and accept oneself, others have to like and accept them first. Gilligan's discussions of Level 2 conventional care were intended to address precisely these sorts of issues. Finally, having already begun with the philosophical discussion above, some further reflection will be undertaken in the final General Discussion chapter to spell out the findings more philosophically and with an eye to some description of a mature ethic of care. 37 C H A P T E R 2: M A P P I N G A N D S C A L E D E V E L O P M E N T STUDIES Introduction The first major goal of Study 1 was to provide a structural clarification and examination of the conventional care construct for which Jack's Silencing the Self Scale (1991) was taken as illustrative in the previous chapter. The circumplex is ideally suited to these ends, given its emphasis upon interpersonal functioning and its empirical success in simplifying a manifold of data into precisely these two orthogonal dimensions of interest (i.e., dominance and nurturance). As a second goal, the process of mapping was continued through the development of additional scales, entitled the Conventional Care Scales (CCS). It was anticipated that the domain of conventional care would occupy a similar region of semantic space as dependency, within Quadrant IV (Friendly-Submissive; see Figure 2 above for a structural representation of agency and communion, otherwise dominance and nurturance). The present study also extends analyses beyond previous self-directed scales to include other interpersonal dimensions hitherto unexamined. The Conventional Care Scales were distilled from a large pool of items reflecting each of three dimensions: self-directed, other-directed, and socially prescribed conventional care. These three dimensions are those frequently employed within interpersonal theory and clinical practice (e.g., Benjamin, 1974, 1996: introject, self-focus, other-focus; Sullivan, 1953). What distinguishes the three dimensions are the different objects towards which expectations of care are directed: towards one's self (self-directed), towards others (other-directed), and the perception of care expectations from others (socially prescribed, see Figure 4 below). The hypothesis that conventional care arises in response to expectations from others to be nice, demure, and silent (i.e., socially prescribed conventional care) is one for which there currently exist no measures. It has only been assumed that the self-directed posture of conventional care arises as the introjected mirror of others. Each dimension was expected to have unique correlates. The notion of mapping, however, was here understood to involve self-directed 38 interpersonal constructs only. Whereas other dimensions can be examined within the Interpersonal Circumplex, the meaning of their projection changes, as is yet to be clarified. Within the self-directed dimension, several constructs added to or clarified current conceptions of the domain of care: anxious concern, self-sacrificial care, and perfectionistic care. The relations among these scales were examined along with other conventional care-like scales, including Jack's Silencing the Self Scale and dependency, employing the Interpersonal Circumplex in its descriptive use as a nomological net. Additionally, a principal components analysis was undertaken to examine the substantive relations among the scale items in an effort to determine the extent of overlap/redundancy among constructs. Such a descriptive clarification was expected to be useful in improving personality and clinical assessment, evaluating the theoretical models behind measures such as the Silencing the Self Scale, and generating linkages to broader conceptual domains such as dependency and attachment constructs. Criteria used for the final selection of Conventional Care Scale items were based on a combination of the more usual substantive criteria (item-total correlations and factor loadings, etc.; Jackson, 1970) and a number of proposed circular structure criteria. Empirical evidence for the use of circular criteria (i.e., alpha-contribution as a function of angular proximity to a scale's central tendency) was examined in the context of the scales used in this study. In short, both substantive (i.e., thematic content) and circular structure criteria were used, both to develop scales as well as to investigate their final relations. Further discussion of structural versus substantive criteria follows. Owing to these additional efforts in test construction, the Conventional Care Scales proved superior to most of the already existing scales in a number of respects. Evidence will be examined suggesting that the two primary facets of conventional care are Octants HI (Unassured-Submissive) and JK (Unassuming-Ingenuous). While an insufficient number of items existed to test final scales, some speculation about the inclusion of a Selfishness and an Intrusive Care scale, reflecting Octants F G (Aloof-Introverted) and NO (Gregarious-Extraverted), respectively (to be 39 described in detail later), among the Conventional Care Scales, will also be examined. The third goal of Study 1 was to extend the process of construct validation across the other three super-factors (i.e., Conscientiousness, Neuroticism, and Openness), as well as to examine their relation with interpersonal capabilities, relationship satisfaction, the false-self, and well-being (i.e., depression and self-esteem). Study 2 replicated the factor structure of the Conventional Care Scales in a second sample. Finally, a third study (reported in a subsequent chapter) involved a secondary analysis of a previous study and examined the comparative predictive validity of conventional care and perfectionism in an eating disordered sample. The preliminary steps involved in the development of the final Conventional Care Scales will first be discussed before examining the final results of Study 1. Construct Explication Stating more specifically what was said in Chapter 1, care, as investigated in the present study, is defined from a "conventional" perspective and fundamentally involves an imbalanced focus on norms as they are instantiated within the complementarity of social roles. Specific measures were developed, expressing the constructs of: anxious concern over being uncaring to others; self-sacrificial care, where the needs of others are considered more important than one's own; and striving to be perfectly caring. Additional preliminary constructs and respective items were identified, suggesting future research directions. While it was argued that many currently existing scales in the care literature have not paid sufficient attention to the dependency literature, the dependency literature would be well served by examining the tacit normative matrix in which relationships operate. While care is a strong social ideal involving concern for others, the qualification of "conventional" is meant to describe both an aspect of intensity (e.g., over/y-nurturant or compliant) and a qualitative interpersonal aspect in reflecting (largely) the discrete friendly-submissive (lower right; see Figure 1 or 2 above) Quadrant IV of the circumplex; intensity/flexibility and circumplex location being the two basic indicators of healthy functioning mentioned in Chapter 1. 40 Provisional Item Development After defining the construct of interest, the next step involved the creation and editing of a pool of items comprising the set from which the final Conventional Care Scales were developed. A provisional set of 137 items was created by item-writers and the author based on a general description of (excessive) care and a review of constructs in the literature. Item writers were composed of a handful of colleagues, lay persons, counsellors, and clinical psychologists. Each was given a sheet of lined paper headed by the construct definition and asked to write at least five items for each of the three dimensions: self-directed, other-directed, and socially prescribed conventional care. A considerably larger sample of self-directed items was included to allow for sufficient items for the series of scales nested within the self-directed dimension. Editing involved the removal of redundant items, correcting for clarity, and rephrasing where necessary to ensure a sufficient number of negative instances. Potential scale items were identified by their thematic content and grouped a priori from among the self-directed items, forming three distinct conventional care scales: Anxious Concern, Self-sacrificial Care, and Perfectionistic Care (see Figure 4). Method Participants Study 1. Participants were 110 male and 192 female undergraduate students drawn from the University of British Columbia, Department of Psychology's subject pool (M = 19.3 years, SD = 1.98). The ethnic composition of the sample was 61.3% Asian (including Chinese, East Indian, Korean, Japanese, Filipino, Vietnamese, Malaysian, and Polynesian), 34.8% White, 2.6% Arabic, and 1.3% unspecified. The gender and ethnic composition of the sample is typical of the department of the university where data were collected. A l l participants received course credit for their involvement. Test administration involved the completion of a questionnaire package composed of twelve randomly ordered self-report measures in addition to a general demographics form. 41 Figure 4 Three Dimensions of Conventional Care and their Respective Scales Perfectionistic Care Self-sacrificial Anxious Care Concern ti ti ti Questionnaire packages were administered in groups, required approximately 75 minutes to complete, and contained the measures described below (see Appendix B). Study 2. The sample of 302 participants in the factor-replicating study was composed of 90 male and 208 female undergraduate students (and four others who failed to indicate their gender and age) drawn from the University of British Columbia, Department of Psychology's subject pool (M = 19.7 years, SD = 2.54). The ethnic composition of the sample was 56.6% Asian, 37.7% White, 1.3% Black, .7% Arabic, .3% each of Hispanic and North American Indian, and 3.0% unspecified. Participants completed a questionnaire 42 containing 87 conventional care items, which required approximately 15 minutes (see Appendix C). Note that all the findings reported below are based on the sample of Study 1, with the exception of the coefficient of congruence statistic reporting the replicability of the factor structure across the samples of both Study 1 and 2. Demographic Information Participants were asked to provide information regarding their age, gender, ethnicity, length of Canadian residency if born outside Canada, year in university, academic major, and average grade (e.g., A+, A , A- , etc.). Information was also requested regarding relationships. If the respondent was currently in an intimate relationship, they were asked to provide its current length and indicate on a 5-point Likert scale (1 = extremely dissatisfied; 3 = neutral; 5 = extremely dissatisfied) their satisfaction and their perception of their partner's satisfaction with the relationship. Factor Structure Measures Revised Interpersonal Adjectives Scales - Big Five (IASR-B5; Wiggins, 1995). The IASR-B5 is an extended version of the Revised Interpersonal Adjective Scales (IASR; Wiggins, Trapnell, & Phillips, 1988). The IASR consists of 64 trait-descriptive adjectives (e.g., "dominant") assessing the eight octants of the Wiggins (1979) circumplex that together are expressive of the coordinates of Dominance and Nurturance. The eight subscales, each reflecting one of the octants, are Assured-Dominant (PA), Arrogant-Calculating (BC), Cold-hearted (DE), Aloof-Introverted (FG), Unassured-Submissive (HI), Unassuming-Ingenuous (JK), Warm-Agreeable (LM), and Gregarious-Extraverted (NO); located relative to the positive pole of the Nurturance axis at 90°, 135°, 180°, 225°, 270°, 315°, 0°, and 45°, respectively. The IASR-B5 supplements the 64-item measure with the inclusion of 20-item markers for each of the three Five-Factor Model factors of Conscientiousness, Neuroticism, and Openness. Wiggins reports alpha coefficients in a large self-report sample of undergraduates 43 ranging from .73 to .86 across subscales. The IASR has been shown to have acceptable stability and validity (Wiggins et al., 1988); and has been used in a variety of studies with both normal and clinical populations (Wiggins, 1995). The IASR is generally viewed as the standard for ideal circumplex criteria. Using an 8-point Likert scale from 1 (Extremely Inaccurate) to 8 (Extremely Accurate), respondents rate each of 124 adjectives for self-descriptive accuracy. An individual circumplex profile can be constructed according to their T-score for each of the eight circumplex scales and, by taking the D O M and NUR factor scores as x and y Cartesian coordinates, the mean length and directionality can also be plotted onto circumplex space. Individuals can then be investigated according to overall profile and/or octant category. Additionally, a score for each of the other three Five-Factor Model factors -Conscientiousness, Neuroticism, and Openness - can be derived. Circumplex Inventory of Interpersonal Problems (IIP-C; Alden, Wiggins, & Pincus, 1990). The IIP-C is a 64-item measure of a variety of interpersonal problems that people report experiencing and which fit around a circumplex about the axes of dominance (DOM) and nurturance (NUR). The longer 127-item Inventory of Interpersonal Problems (IIP; Horowitz, Rosenberg, Baer, Ureno, & Villsefior, 1988) was revised to circumplex criteria by Alden et al. (IIP-C) and has been demonstrated to structurally converge to the Revised Interpersonal Adjective Scale (IASR; Wiggins et al., 1988). The IIP-C is divided into two sections according to the type of items included: "things you find it hard to do with other people" and "things that you do too much." Participants rate each of the 64 statements on a 5-point Likert scale from 0 (Not at all) to 4 (Extremely), yielding eight octant raw-scores varying from 0 to 32. Each octant is represented by four items for each of the two types, for a total of eight items per octant. The eight subscales, each reflecting one of the octants, are Domineering (PA), Vindictive (BC), Cold (DE), Socially Avoidant (FG), Nonassertive (HI), Exploitable (JK), Overly Nurturant (LM), and Intrusive (NO); located relative to the positive pole of the x or NUR axis at 90°, 135°, 180°, 225°, 270°, 315°, 0°, and 45°, respectively. 44 The authors report alpha coefficients across the eight scales ranging from .72 to .85. The IIP has demonstrated test-retest reliabilities across a 10-week interval ranging from .80 to .90, and amongst a clinical sample demonstrated sensitivity to change as a result of psychotherapy, and was meaningfully related to other concomitant measures of psychopathology (Horowitz et al., 1988). The IIP-C has been used by a number of investigators to examine individual differences in interpersonal style and has been shown to have good structural convergence across self- and peer-ratings (e.g., Wagner, Kiesler, & Schmidt, 1995). Dependency and Conventional Care Measures Irrational Beliefs Test - Demand for Approval Subscale (IBT; Jones, 1969). The IBT - Demand for Approval Subscale measures the extent to which individuals assert 10 of the core irrational beliefs described by Albert Ellis, the originator of Rational Emotive Therapy. Factor analyses have confirmed the structure of the total measure (Lohr & Bonge, 1981) and test-retest reliability over an 8-week period has demonstrated considerable temporal consistency. In the present study, the alpha coefficient for this scale was .78. The measure has been shown to predict a number of clinical problems including depression (Nelson, 1977) and anxiety (Lohr & Bonge, 1981). Nelson (1977) found gender differences for the relationship between the Demand For Approval Subscale and depression, in favor of women showing the stronger relationship (z - 2.55, p < .05). The current study used only the 12-item Demand for Approval Subscale (e.g., "I want everyone to like me") which correlates .80 with the total scale. Participants rated their item agreement on a 5-point Likert scale from 0 (Disagree) to 4 (Agree), yielding a range from 0 to 60. High values indicate a high need for approval from others. Conventional Care Scales (CCS). Participants rated each of 137 provisional items on a 7-point Likert scale from 1 (Strongly Disagree) to 7 (Strongly Agree). A combination of circular and traditional criteria (Jackson, 1970) was used to determine final item inclusion in their respective theoretically derived scales. Items from the provisional pool were dropped 45 if: (a) they were highly skewed (i.e., less than 7% of endorsements were located in three adjacent rating points on the 7-point scale), (b) they were undiscriminating (i.e., standard deviation less than 1), (c) their corrected item-total correlations were less than .40 on their respective subscales as defined a priori by thematic content, and (d) in a final principal components factor analysis, they loaded less than .40 on their respective factor, and did not load above .40 on more than one factor. Criteria for item inclusion within the three self-directed Conventional Care Scales involved additional circular criteria to establish the circular meaningfulness of the construct. Items were dropped if their angular projections were outside their target octant (i.e., octant midpoint ± 30°). A slightly broader range beyond the bounds of the octant was used so as to maintain a balance between constructs that were octant-defining, taking the octant as next superordinate level, and not unduly constraining the item range for those constructs whose circular mean was nearer the octant boundary than the center point. Those eight items with the highest factor loadings were finally retained. The circular factor loading was defined as (see Equation 1 below for more detail): V. * cos(0, - </>) (1) where <p. is the angle (in radians) defining the midpoint of the item's octant category (i.e., 270° and 315° for octants EQ and JK, respectively) and 9. is the angle of projection for item i. In total, the Conventional Care Scales are composed of five scales; three of which are self-directed and two of which are not (one other-directed and one socially prescribed). Reliabilities and further descriptive statistics for the Conventional Care Scales are reported in the subsequent section, "Conventional Care Scale Development." Silencing the Self Scale (STSS; Jack, 1991). The STSS is composed of four subscales derived from the reports of depressed women regarding disturbances in their sense of self. Only the directly interpersonal Care, Self-silencing, and Externalized Self-perception Subscales were used in the current study. Additionally it was thought that more highly validated versions (e.g., Campbell, 1990) of the divided-self construct were available and 46 thus would make less of a contribution to empirical research than the other three subscales. Theoretically, the STSS was informed by Gilligan's (1982) theory and attachment theory (e.g., Bowlby, 1977). Care is a measure of the extent to which "goodness" is understood as self-sacrificial and placed in the service of "maintaining relationships" (e.g., "Caring means putting the other person's needs in front of my own"). The Self-silencing subscale, hereafter Silence (reserving Self-silencing for the total value across scales), measures the tendency to inhibit self-expression in order to avoid interpersonal conflict (e.g., "I don't speak my feelings to my friends when I know they will cause disagreement"). Externalized Self-perception, hereafter simply External, reflects the degree to which respondents judge themselves from the perspective of others (e.g., "I tend to judge myself by how I think my friends see me"). Across several samples - battered women in shelters, female college students, and cocaine-involved mothers - alphas have ranged from .86 to .94 and test-retest reliabilities ranged from .88 to .93 (Jack & Di l l , 1992). In the present study, the alpha coefficients were .64, .83, and .71 for the Care, Silence, and External Subscales, respectively. Principal components analyses have been conducted demonstrating that items load satisfactorily on their respective subscales. Participants rate their level of agreement on a 5-point Likert scale from 1 (Strongly Disagree) to 5 (Strongly Agree). High values indicate problems with self in relation to others. Unmitigated Communion Scale (UCS; Helgeson & Fritz, 1998). The UCS is a 9-item measure, broadly conceptualized as "a focus on others to the exclusion of the self" which is reflected in items like placing others' needs before one's own (e.g., "I always place the needs of others above my own") and an excessive concern with others (e.g., "I worry about how other people get along without me when I am not there"). Across several studies alpha reliabilities have ranged from .7 to .8 and very good test-retest reliabilities have been reported (Helgeson, 1993, 1994). In the present study, the alpha coefficient for this scale was .68. 47 Respondents indicate agreement or disagreement with each of the items on a 5-point Likert scale from 1 (Strongly Disagree) to 5 (Strongly Agree). A single scale score is calculated summing across items, with high scores indicating an excessive care focus on others. Social Perfectionism Scale (SPS; Wiebe & McCabe, 1998). The SPS seeks to explicitly tap the domain of expectations and standards for social interactions and relationships. Two dimensions, self-directed and other-directed, define each of two subscales (SPS-SD and SPS-OD, respectively), composed of 10 items each. A representative self-directed item reads, "It's important that I say all the right things in a conversation," and a representative other-directed item reads, "To be worthy of my friendship, others should live up to my expectations." The authors do not report reliabilities; however in the current study the alpha coefficients were .77 and .74 for the Self- and Other-directed Social Perfectionism Subscales, respectively. Respondents rate their level of item agreement on a 7-point Likert scale from 1 (Very Strongly Disagree) to 7 (Very Strongly Agree). High scores indicate high levels of social • perfectionism for each of the respective subscales. Outcome Measures In addition to comparing and evaluating conventional care and dependency scales in relation to the Interpersonal Circumplex, here employed as a nomological net, measures of depression and self-esteem were administered to determine level of adjustment. Clinical scales should predict maladjustment over and above widely employed general indices of functioning. Social desirability was also assessed to determine the extent to which the scales developed in the current study represented an accurate reflection of the construct rather than an individual difference in style of social responding. Finally, a mini-narrative method was employed as a means of examining the interaction between the identification of oneself with caring and experiences of rejection in that role. A negative affect scale was used by participants to evaluate the experience about which they wrote. 48 Marlowe-Crowne Social Desirability Scale (MC(10); Strahan & Gerbasi, 1972). The MC(IO), hereafter simply M C , is a short 10-item true-false self-report measure intended to assess the extent to which respondents present or attribute unrealistically favorable characteristics to themselves (e.g., "I'm always willing to admit it when I make a mistake"). The M C has shown moderate internal consistency and is highly correlated with the longer 25-item Marlowe-Crowne Social Desirability Scale (Crowne & Marlowe, 1960). Scores on this measure range from 0 to 10. High values indicate a tendency to present oneself in an unrealistically favorable light. In the present study, the alpha coefficient for this scale was .44. Rosenberg Self-esteem Scale (RSES; Rosenberg, 1965). The RSES is a global self-report measure of one's sense of self-worth. General items allow respondents to select and evaluate their own relevant content-specific behaviors as the bases for judgment (e.g., "I feel that I am a person of worth, at least on an equal basis with others"). The measure is widely used and demonstrates good psychometric properties (Wylie, 1974); including a test-retest coefficient of .85 (Silber & Tippett, 1965) using a college sample, and a high internal consistency (a = .78; DeLongis, Folkman, & Lazarus, 1988). In the present study, the alpha coefficient for this scale was .89. Participants rated each of 10 statements on a 4-point Likert scale from 1 (Strongly Agree) to 4 (Strongly Disagree), yielding a range from 10 to 40. High values indicate a high level of self-esteem. Beck's Depression Inventory (BDI; Beck, 1967). The BDI is a 21-item instrument measuring the behavioral, cognitive, motivational, and somatic symptoms of depression. The scale measures state ("... the way you have been feeling the PAST WEEK, INCLUDING TODAY", as opposed to trait aspects of the depressive phenomenon; Beck & Beamsdorfer, 1974), and is widely used as a measure of severity (vs. chronicity) of current depressive symptoms. The reliability and validity of the measure have been demonstrated across a number of studies (Beck, Steer, & Garbin, 1988, for a review). In the present study, the alpha coefficient for this scale was .87. 49 Respondents select one of four self-descriptive statements, each of which is more reflective of the severity of clinical depression than the previous. An overall score is derived by summing across items, with the exception of a weight-loss item which is included only on the condition that respondents circle "no" to the question, "I am purposely trying to lose weight by eating less." General level of depression can be indicated by cut-off values: severe (24 and above), moderate (16-23), mild (10-15), and absence of depression (0-9). Interpersonal Competencies Questionnaire (ICQ; Buhrmester, Furman, Wittenberg, & Reis, 1988). The ICQ is a 40-item questionnaire which assesses five rationally and factor-analytically derived domains of interpersonal competence. The three considered most relevant for this study are: Negative Assertion (ICQ-NA; e.g., "Telling a companion you don't like a certain way he or she has been treating you"), Self-disclosure (ICQ-SD; e.g., "Telling a close companion about the things that secretly make you feel anxious or afraid"), and Emotional Support (ICQ-ES; e.g., "Being a good and sensitive listener for a companion who is upset"). The authors report Cronbach alphas for each of the five subscales ranging from .77 to .86, and moderate relations between the ratings of self and others. In the present study, the alpha coefficients for the Negative Assertion, Self-disclosure, and Emotional Support Subscales were .86, .80, and .86, respectively. Construct validity has been demonstrated across a variety of measures. For example, satisfaction in new relationships was best predicted by initiating competence, whereas satisfaction in friendships is best predicted by emotional support. Participants rate the degree of competence and comfort they would experience for each interpersonal situation using a 5-point Likert scale, yielding a subscale range of 8 to 40: 1 = I'm poor at this; I'd feel so uncomfortable and unable to handle this situation, I'd avoid it if possible. 2 = I'm only fair at this; I'd feel uncomfortable and would have lots of difficulty handling this situation. 3 = I'm OK at this; I'd feel somewhat uncomfortable and have some difficulty handling this situation. 50 4 = I'm good at this; I'd feel quite comfortable and able to handle this situation. 5 = I'm extremely good at this; I'd feel very comfortable and could handle this situation very well". High scores indicate reported competence within the respective domain. Positive and Negative Affect Scales - Negative Affectivity Subscale (PANAS: Watson, Clark, & Tellegen, 1988). The PANAS is a measure of dispositional negative affectivity (NA; e.g., nervous, hostile, ashamed) and positive affectivity (PA; e.g., proud, enthusiastic, inspired). N A and PA have been identified by Watson and Tellegen (1985) as the most superordinate dimensions of emotional experience. Because N A is strongly related to the Big-Five Neuroticism factor, and PA is moderately related to the Big-Five Extraversion factor (Meyer & Shack, 1989; Watson & Clark, 1992), some investigators have made the argument that Neuroticism and Extraversion be re-labelled "Negative Emotionality" and "Positive Emotionality," respectively (Tellegen, 1985; Watson & Clark, 1984). Each of the scales contains 10 mood descriptors which have been shown to be orthogonal. Alpha reliabilities are acceptably high, ranging from .84 to .87 for NA. Test-retest reliabilities over an 8-week interval have shown acceptable stability, ranging from .39 to .71 depending upon the instructed time frame (i.e., "right now" to "in general"). Participants in this study were asked to write two narratives of care, completing the Negative Affectivity Scale after each. In the first narrative participants were asked to, "write a story about a time when a friend (or partner) either ignored or refused your help or care." The second narrative involved writing "a story about a time when a friend (or partner) went to someone other than yourself'to get and receive assistance or care." Participants then described how the event made them feel across the 10 N A descriptors. Two additional descriptors, "sad" and "disappointed," were included, in addition to a single question regarding the extent to which the incident caused them to ruminate. A l l responses were indicated on a 5-point Likert scale from 1 (Very slightly or not at all) to 5 (Extremely). A total score was calculated by summing across all items, including the two additional 51 descriptors and the rumination item (i.e., 13 items in total). High values indicate considerable distress as a result of their either not being sought out (i.e., implicitly refused) as a care-giver or being explicitly refused as a care-giver. Low values indicate calmness or serenity and an absence of rumination. Results Owing to the detail of the following analyses, a more extended effort is taken below to provide a guiding framework for the reader. We begin by examining the dependency and conventional care constructs within the space of the Interpersonal Circumplex, a "super" circular space which will be described in detail in the initial sections below. Next, circular item analyses were undertaken by projecting individual scale items of both conventional care and dependency constructs into circular space. The provisional pool of self-directed items was also projected into circular space. The guiding principle informing such an analysis is summarized in the first of two circumplex propositions: that angular proximity within circular space reflects the degree of similarity in their underlying interpersonal functioning. It was hypothesized that current scale items would span an excessively broad array of interpersonal space, oversampling from an array of semantic meanings that are best kept distinct. It is argued that both conventional care and dependency scales are multifaceted, which in the context of the circumplex means that they fall within more than one octant reflecting meaningful interpersonal differences. Recall that meaningful distinctions within the circumplex exist at the level of octants, if not more narrowly at 16s, etc. Put elsewise, not only is there more than one way to be scored at conventional Stage 3 in Kohlberg's scheme, but more specifically there is more than one way to be found on the obligations (vs. entitlement) side of the imbalance. This study sought to correct this defect by (a) identifying scales that possess adequate circular structure and substantive characteristics, and (b) developing additional scales incorporating both criteria into their development. 52 In the subsequent sections, the rationale for the inclusion of circular criteria within more traditional procedures of test development are examined. Three self-directed Conventional Care Scales were developed based on the conclusions here. Procedures and descriptive data regarding these scales are also reported. The ideal was held out in this dissertation that both circular structure and substantive approaches would result in the most psychometrically sound self-directed scales which are at the same time the most meaningful. The following sections proceed in detail, octant by octant, seeking to verify the psychometric utility of circular structure criteria, specifically angular location, by examining (a) scale alpha-contribution plots and (b) substantive factors arising from a principal components analysis of all scale items. The broader purpose was to assess which of the various scales were redundant/overlapping, most theoretically meaningful, and psychometrically sound. Then, in a section entitled "Beyond the Self-directed Dimension," the development of the non-self-directed Conventional Care Scales (i.e., other-directed and socially prescribed) is discussed. The development and discussion of these non-self-directed scales are undertaken after the self-directed circumplex analysis, given the limited utility of such analyses for non-self-directed scales. The significance and meaningfulness of circular analyses are given credence by the linkage that the various octants have with previous studies employing the Interpersonal Circumplex as a guiding structure or nomological net. A broader discussion is undertaken examining some of the interpersonal dynamics reflective of the respective octants. This will again be undertaken octant by octant. Continuing the process of construct validation, the relation of conventional care and dependency scales (including those not developed in the current study) with interpersonal competencies is then examined. Recall that Jack (1991) argues that care and silencing are for the purpose of maintaining relationships. It will be argued that there exists also a distinctly submissive-avoidant form of conventional care, or perhaps "concern" would be a better 53 descriptor, leaving open the possibility that care (or concern) has less to do with the other than preservation of, or concern for, the self. Analyses of the conventional care and dependency scales is then extended to the remaining three factors of the Five-Factor Model: Conscientiousness, Neuroticism, and Openness. While scales with similar interpersonal patterning may underlie scales falling within one and the same octant, they can at the same time be markedly different, depending on the degree of "coloration" they are given from other factors. In the final sections, gender differences are examined and a series of hierarchical regression analyses is undertaken examining the incremental validity of scales' prediction of adjustment (i.e., depression and low self-esteem). First, we turn to a description of the creation of a "super" circumplex space which serves as the platform for much of what follows. Creating the "Super" Circumplex Space A "super" circumplex factor space was constructed by jointly factoring two Interpersonal Circumplexes: the eight octant scales of the IASR and the eight octant scales of the IIP-C (Pincus & Gurtman, 1995). A preliminary conception of a "super" factor space can be grasped by imagining the imposition of the IIP-C and the IASR atop one another (see Figure 5 below). Octant JK (Unassuming-Ingenuous) is highlighted with a projected item from each scale marking the octant. The IIP-C item reflects a "problems" version of the IASR interpersonal trait. More will be said regarding the relation of the two scales for their respective octant. First, some explanation of the steps in the construction of the super circumplex or circle are discussed. Of the two measures constituting the circumplex factor space, the IIP-C first requires prior ipsatizing because of the strong intercorrelation that exists between subscales. The strong relation between scales results from a general tendency for scores across all items to either rise or lower systematically relative to an individual's tendency to report distress; referred to as a general "complaint" factor. This individual tendency results in an inflated 54 F i g u r e 5 A Comparison of IASR and IIP-C Scale Items from Octant JK (Unassuming-Ingenuous) within Super Circumplex Space (DOM + ) (301°) correlation among the subscales and can inaccurately produce higher or lower correlations with concomitant measures, as these correlate with the complaint factor. Correlational analyses, thus, require ipsatization across the 64 circumplex items to remove this first Complaint factor. An ipsative IIP-C scale score represents a z-score deviation from the individual's own mean response to all 64 items of the inventory, showing whether a particular score reflects higher-than-average or lower-than-average distress for that person. Removal of the complaint factor by ipsatizing scores results in subscales which are relatively independent 55 of each other and follow circumplex criteria (i.e., scale scores closer to one another are more highly correlated than those further away, up to 90° which are then orthogonal; see Circumplex Proposition 1 below). Having ipsatized the IIP-C, a principal components analysis can then be conducted upon the final 16 scales (i.e., IASR and IIP-C combined), extracting two large interpersonal factors (eigenvalues 5.04 and 4.31; the third and fourth factors had eigenvalues of 1.23 and .88, respectively). The two extracted principal components accounted for 58.4% of the variance in the sample. A dominance and nurturance coordinate system in this two-factor space was created by rotating the two extracted principal components from the initial solution until the obtained angles for the 16 octant scales are in maximum convergence with the theoretical expectations as specified by the circumplex. Angles are derived from the factor loadings. Since it is unlikely that a standard varimax rotation will ensure the ideal circumplex rotation (i.e., L M = 0°; NO = 45°; PA = 90°, etc.), a Procrustes rotation was used to provide a transformation matrix that can be rotated to a least-squares fit to any fully specified target matrix. (A varimax rotation will not converge in a perfect circumplex since any rotation is as good as any other by simple structure criteria.) The target of rotation was the following matrix of cosines and sines (Wiggins & Broughton, 1991, p. 349): PA B C D E F G HI JK L M NO NUR .00 -.707 -1.00 -.707 .00 .707 1.00 .707 D O M 1.00 .707 .00 -.707 -1.00 -.707 .00 .707 The respective values assigned to first row above are the target loadings for the first factor of Nurturance (NUR), and those assigned to the second row are the target loadings for the orthogonal Dominance (DOM) factor. As indicated in Figure 6, the rotated axes (i.e., the midpoints of the representative octants in the figure) fall between the octant scales for the two measures, the IASR and the IIP-C. In Octant PA (Assured-Dominant) for example, the IASR (PA) Scale is located at 85° 56 Figure 6 Structure of the IASR and the IIP-C in "Super" Circumplex Space (DOM + ) (PA) Assured-Dominant (DE) (HI) Unassured-Submissive Note. The solid boxes indicate the IIP-C and the open boxes the IASR. and the IIP-C (PA) Scale is located at 103° - with respect to the right side of the x-axis (NUR+) marking 0°. The midpoint of Octant PA in the "super" circumplex space (90°) falls between the IASR and IIP-C octant scales. In Octant B C (Arrogant-Calculating), the IASR and IIP-C octant scales fall virtually on top of one another. As a series of sine-cosines the IASR in general lags with respect to phase, or put elsewise, the IASR scales are generally located clockwise to IIP-C. The super circumplex space is understood as a distinct superordinate space marked by the various scales of the IASR and the IIP-C. While a number of studies have involved using either the IASR or the IIP-C as nomological nets, only one prior study to the current one has employed a "super" space - which while requiring 57 twice as many scale items to construct has the theoretical advantage of combining investigations in the field. As Gurtman (1993) notes, construct locations are not necessarily interchangeable in different circumplex measures and their spaces. For example, although both the IIP-C and Wiggins's (1979) IAS have excellent circumplex properties, projections of variables are likely to be slightly off, perhaps by about 15° on average, (p. 259) Gurtman's conclusion based on different studies is confirmed in the present study where both the IIP-C and the IASR are considered concurrently in the same space. The two scales quite clearly relate to one another as anticipated, with scales falling within their respective octants, but not falling on top of one another. Projecting Outside Variables onto Circular Space Some researchers seeking to examine outside clinical variables have chosen the IIP-C because it can be considered a clinical or "problems" version of the IASR, but there is no reason that the two can not be considered together in the interest of joining clinical with normal personality research. In addition to increasing the reliability of projecting outside variables into circular space, a "super" space can provide a vantage point or single index by which constructs of varying sorts within the field can be compared. The current use of a "super" space takes its lead from Pincus and Gurtman (1995) who first proposed the idea. "Projection" of the outside variable (or component item) onto the newly created "super" space is the second major step. Any outside Variable X can be located on the D O M and NUR plane by regarding the variable's correlation with the extracted D O M and NUR principal component (i.e., factor) scores as Cartesian coordinates. Individual factor scores are obtained by premultiplying the transformation matrix (obtained in the process of matrix manipulations involved in the Procrustes rotation) by the initial factor scores saved from the principal components analysis. Outside Variable X can then be correlated with the D O M and NUR factor scores for use in the trigonometric procedures described in Figure 7. The Pearson product-moment correlation between D O M and Variable X provides the distance 58 Figure 7 Projection of an Outside Variable onto an Interpersonal Space DOM + + 1.00 f \ r / / y i /9 -1.00 + 1.00 -LOO ^^y^ NUR+ Note. Adapted from "A Geometric Taxonomy of Personality Scales" by J. Wiggins & R. Broughton, 1991, European Journal of Personality, 5, p. 348. Copyright 1991 by Wiley. from the origin of the variable on the horizontal axis (x = RDOM variable x)> t n e correlation between NUR and Variable X locates the variable on the vertical axis (y). The variable's angular displacement onto the circumplex would then be given by the geometric formula 6 = tan-](y/x), and the length of the associated vector by r = (x2 + y2)Vl (i.e., the square root of its communality with the D O M and NUR factors). The angle score (ff) refers to the summary "direction" or tendency in interpersonal space, taking the positive pole of the NUR axis as its referent. Vector length (r) refers to the magnitude or extent of that interpersonal tendency. Locating the resultant point provides an overall summary of an individual's or group's interpersonal behavior plotted within the circumplex along the dimensions of: (a) "intention to influence, change, or control other" (DOM; dominance to submissiveness), (b) "nature of 59 involvement" (NUR; friendly or hostile), in addition to the (c) "degree of psychological involvement" (Horowitz, 1979) reflecting the "intensity" of a particular interpersonal behavior or more germane to the current study, the extent to which the projected Variable X is considered interpersonal in nature. For example (see Figure 5 above), the IASR interpersonal trait "Undemanding" is located at 295° in super space while the IIP-C problems version, "It is hard for me to confront other people with problems that come up," is located at 301°. (Comparison of Octant JK Unassuming-Ingenuous items anticipates the focus of attention below.) The thematic relation is clear, the IIP-C is a "problems" version of the IASR. To grasp the relation between projected variables, including the IASR and IIP-C octant scales which form the factor markers of the super circumplex, some explanation is necessary. Circumplex Proposition 1 (formalized from Gurtman, 1993) states: Angular proximity within the circle is a direct measure of the similarity of constructs in interpersonal content. One of the attractions of the circumplex is its amenability to the use of relatively simple geometric circular statistics. In an ideal circumplex with scales of unit vector length the correlation between projected items (and scales) occupying the same location on the circle would be 1. Generally, items falling onto different locations of the circle would correlate according to the cosine of the angle of separation (in radians) between the two variables (Equation 2): r = cos(0,-fil) (2) Items at 45° to one another would correlate .707, and the correlation of scales separated by 90° would be 0. (The cosine of 0° is 1, whereas the cosine of 90° is 0.) The correlation of scales at 180° would be -1. Saying that an item falls within a particular location along the circle is a comment about its relation not only with the octant in which an item falls (i.e., correlates maximally) but is also a comment about its relation with all the other points or octants in the circle. Plotting a variable within the circle is equivalent to examining a series 60 of correlations with all points (e.g., octants) making up the circle, and whose pattern will form a sinusoidal curve or more specifically, a cosine function. It is in this latter "gestalt" sense that the circumplex is understood as a nomological net (Gurtman, 1992). Variables, however, are never completely described, nor measured without error by two factors. Whether the two factors are dominance and nurturance or any two others, resultant vector lengths will thus never reach unit length (i.e., vector length = 1). For example, scales within the IASR alone or the IIP-C alone follow very close to this circular ideal as indicated in Circumplex Proposition 1, but within the super circumplex these two scales are something of "substantively" nested apples-and-oranges sharing similar underlying (or superordinate) interpersonal characteristics. They are factor "markers" of the same slice of interpersonal space. The correlation between IASR (BA) and IIP-C (BA) is moderate (r = .42), with respective octant scores across the two circumplex systems varying (rs = .22 for Octant D E to .63 for PA; see Figure 6). A second qualifier is necessary when examining the projections of outside variables, be they within single circumplex spaces such as the IASR or IIP-C, or a super circumplex as constructed here. "Substantively different constructs (e.g., competitiveness, resentfulness) can occupy similar locations on the circle" (Gurtman, 1993, p. 257). In the super circumplex, the individual scales of the IASR and the IIP-C become narrower hierarchically nested constructs within the broader and hierarchically superordinate octant in which they fall. Circumplex Proposition 2 states, Projections will approach the circular ideal (Circumplex Proposition 1) to the extent that items share the same substantive or thematic interpersonal content. Projections within the same octant can possess unique variance accorded narrower constructs yet share with one another a broader correspondence in their (underlying/superordinate) interpersonal structure. The working out of this in what follows, in assessing and developing scales, will make the above points clearer. Based on Circumplex Proposition 1, angular proximity within the circle is a direct measure of the similarity of constructs in interpersonal content, it follows that, in addition to 61 examining scales in their nomological relation to one another, a scale's items can also be examined with respect to their fidelity to the theoretical construct of which they are intended as a componential operationalization. That a circular item analysis of this sort is ideally conducted amongst items reflective of the same construct is expressive of Circumplex Proposition 2. The use of both circumplex propositions expresses the importance of substantive considerations as well as structural ones. Purely substantive approaches ... do not guide the test developer in the selection of appropriate constructs to measure.... On the other hand, purely structural approaches may not be adequate to ensure substantive specificity of measures. (Gurtman, 1993, p. 257) Put elsewise, while structural criteria alone - and this is also true of any structural approach including the Five-Factor Model (Hogan, 1978; Wiggins & Broughton, 1985) - can not ensure item goodness, "structural approaches can identify when measures (or items) are substantively incorrect" (Gurtman, 1993, p. 257). By these lights the current study sought to assess a number of constructs that can generally be regarded as a form of "care gone awry." A number of additional measures were constructed, guided both by the care literature and the a priori formation of constructs generated by item-writers. A comparison and contrast of scales was undertaken using the super circumplex in its role as a descriptive nomological net. Additional measures were constructed based on a proposed set of circular criteria (for the individual criteria see above under "Dependency and Care Measures" Conventional Care Scales, details of which will follow) which were used in conjunction with more traditional approaches to test development. The theoretical underpinnings of the "substance plus (circular) structure" approach were outlined above, but their empirical use/advantage have yet to be demonstrated. Circular Analysis of Scale Items While considerable consensus exists that dependency is a multifaceted construct, the basis for making distinctions remains controversial. Pincus and Gurtman (1995) conducted a 62 structural analysis of dependency measures using the Interpersonal Circumplex as a nomological net to compare and assess the grounds upon which the various multifaceted models of dependency were being proposed. A number of widely used measures of dependency, including the Sociotropy-Autonomy Scale (Beck et al., 1983), Interpersonal Dependency Inventory (Hirschfeld, Klerman, Gough, Barrett, Korchin, & Chodoff, 1977), Depressive Experiences Questionnaire (Blatt, D'Afflitti, & Quinlan, 1976), MMPI -Dependent Personality Disorder Scale (Morey, Waugh, & Blashfield, 1985), and the Dysfunctional Attitudes Scale (Weisman & Beck, 1978), were projected onto interpersonal space as outlined above. Individual scales varied in the extent to which they sampled across interpersonal space. While the current study had the space to include only a single measure of dependency (i.e., Irrational Beliefs Test - Demand for Approval Subscale; Jones, 1969), it possessed a similarly broad distribution of its 12 test items (see Figure 8). Most items, however, fell into the HI (Unassured-Submissive) octant. The present study pursued a similar strategy, and with similar results, to that of Pincus and Gurtman (1995) in their investigation of the dependency construct. Figure 9 shows the projection of items from the Silencing the Self Scale (STSS), the Unmitigated Communion Scale (UCS), and the self-directed dimension of the Social Perfectionism Scale (SPS-SD) onto the surface of interpersonal space. Although employing fewer items, the resulting broad circular dispersion appears very similar to the projection of dependency scale items in the Pincus and Gurtman (1995) study. Figure 10 repeats the analysis for conventional care items generated by item-writers and a review of the literature. While the construct of dependency has thus far been described as a component of conventional care, and vice versa, analyses have been kept separate for comparative purposes. That the circular space that they all inhabit overlaps is clear. Items within what has here been broadly termed (self-directed) conventional care, rather than being narrowly confined, spanned a distribution of over 180°. Items generated by item-writers and a literature review (i.e., the provisional item pool) spanned an even greater 63 Figure 8 Circular Dispersion of Items for the Irrational Beliefs Test - Demand for Approval (dependency) Subscale (DOM + ) (PA) Assured-Dominant (DE) (HI) Unassured-Submissive range, with the densest distribution occupying a 90° arc defining the friendly-submissive quadrant of the circle (i.e., Octants HI, JK, L M ) . These circular analyses suggest that conventional care, like dependency, is an interpersonally diverse trait. This conclusion derives from the characteristics of the circumplex itself. The wide dispersion of dependency and conventional care items represents a large pool from which the various scales selectively or broadly sample from an array of interpersonal meanings. Do purportedly distinct measures of the various interpersonal facets of dependency in fact possess a consistent set of items falling within a relatively narrow interpersonal space not much wider than a single octant? 64 Figure 9 Circular Dispersion of Items for the STSS, UCS, and the SPS-SD scales (DOM + ) (PA) Assured-Dominant (DE) (HI) Unassured-Submissive A single scale which possesses a distribution of items crossing 90° contains interpersonal contents which are orthogonal to one another. Circumplex Proposition 1 makes the empirical point. Its theoretical counterpart is supported in the findings of meaningful differences not only at the level of octants, but also at sixteenths (Gifford & O'Connor, 1987). Scale developers using substantive criteria alone may include items that correlate with one another on the basis of some additional factor such as shared distress (i.e., Neuroticism), but tapping an excessively broad array of interpersonal meanings will result in a lack of theoretical clarity. The addition of circular structural criteria is meant to help test developers create more theoretically meaningful scales. 65 Figure 10 Circular Dispersion of the Provisional Pool of Self-directed Conventional Care Items (DOM + ) (PA) Assured-Dominant (HI) Unassured-Submissive To conduct scale analyses two indices of item variability were calculated. For each scale a circular mean (CM; ci. Mardia, 1972; Pincus & Gurtman, 1995) was computed as the angular location (in radians) within the circle (angle (j)) that minimizes the value: I V , * cos(0. - ^) (3) where V; is item l's vector length (i.e., the square root of its communality with the dominance and nurturance principal components), and 61 is the item's projected angular location on the circle. (The circular mean closely corresponds to the circular location as more standardly found by projecting the total scale onto the circle.) From Equation 3, a second index of item variability, hereafter referred to as the standardized circular dispersion (SCD), can be 66 computed from the mean item loading in Equation 3: 1/k * SV; * cos(0. - <p) (4) where k is the number of items in the scale. The higher, then, is the resultant value, the greater will be the average loading of items on the factor (vector) that expresses the mean. The standardized circular dispersion, a variant of Mardia's (1972) circular standard deviation, is then calculated by dividing the result of Equation 4 by the mean vector length of the items (Ilk *. E V ) . The mean item vector length is simply the average of the sum of item i's vector length. In the analysis of Pincus and Gurtman (1995), the desirably narrow culling of items by test developers was found to vary. For example, while items within Beck's Sociotropy-Autonomy - Pleasing Others Subscale were narrowly confined within Octant JK Unassuming-Ingenuous (CM = 311°, Octant JK, SCD = 12.3°), the Interpersonal Dependency Inventory - Emotional Reliance Subscale items were very broadly distributed (CM = 324°, Octant JK, SCD = 79.3°). The other purported measures of dependency which Pincus and Gurtman report fell somewhere between these two (M = 34.1°, calculated from reported data) - where the mean here refers to the average standardized circular distribution (SCD) across all scales. Shown in Table 2 is an iconic depiction of the item distribution for the two above scales based on the data reported in Pincus and Gurtman (1995), along with the Irrational Beliefs Test - Demand for Approval Subscale (Jones, 1969) employed in the current study. Comparing the circular dispersion of items from the Demand for Approval Subscale (see Figure 8 above) with the standardized circular dispersion (SCD) indicates that the SCD encompasses the central cluster of eight of the twelve items. The purpose of the SCD is to provide some standardized indication of dispersion by which scales can be compared. Examining the iconic representations of conventional care scales developed outside the current study indicates a similar divergent sampling of interpersonal meaning. The SCD of the Unmitigated Communion Scale reveals a considerable circular distribution of items (SCD = 35.3°; see Table 3). The Silence Subscale possessed the narrowest sampling of 67 Table 2 Comparison of Scale Circular Statistics for Dependency Measures Scale CM SCD Mean vector length Mean item loading Pleasing Others (SAS) Q 311° 12.3° .28 .28 Emotional Reliance (IDI) 324° 79.3° .13 .02 Demand for Approval (IBT) 271° 34.4° .22 .18 Note. Data for the Sociotropy-Autonomy Scale - Pleasing Others (Beck et al., 1983) and the Interpersonal Dependency Inventory - Emotional Reliance (Hirschfeld et al., 1977) subscales are from Pincus and Gurtman (1995). Data for the Irrational Beliefs Test - Demand for Approval Subscale (Jones, 1969) are from the current study. meanings (SCD = 14.8°). The SCD, however, especially for scales with as few items as these care/dependency scales, will not be a good indicator of outlying items which will be examined in more detail in analyses that follow. Previous studies projecting final scales onto interpersonal space will clearly not reveal the entire story in those cases where individual items are actually tapping a broad array of interpersonal tendencies. Thus, similar to several of the dependency scales analyzed by Pincus and Gurtman, some conventional care scales contained items separated by 90° and more, reflecting interpersonal traits orthogonal to one another. Thus far, details regarding the creation of a "super" circumplex space from two interpersonal measures, the IASR and the IIP-C, have been discussed. It was proposed that an analysis of projected items, rather than just total scale scores reflecting the majority of 68 Table 3 Comparison of Scale Circular Statistics for Measures of Conventional Care Scale CM SCD Mean vector length Mean item loading Care (STSS) Q 312° 24.6° .15 .14 Silence (STSS) 253° 14.8° .34 .33 External (STSS) 251° 14.9° .22 .22 Unmitigated Communion (UCS) G 338° 35.3° .25 .21 Self-directed Social Perfectionism (SPS) Q 262° 28.0° .22 .20 work in the field prior to this dissertation, would indicate a degree of diversity otherwise masked. Hypotheses regarding item analyses were confirmed, care scales along with the provisional pool of self-directed conventional care items revealed a broad array of items similar to the Pincus and Gurtman analyses of dependency. Items within one and the same scale reflect an overly broad (even orthogonal) array of interpersonal meanings. Arguments were presented based on the first of two circumplex propositions as to why this represented a problem. In contrast, more analytic approaches which include circular item-proximity criteria should evidence both (a) improved psychometric characteristics and (b) greater theoretic 69 clarity. Evidence for the first point will be garnered in the following section by examining the alpha-contribution plot of current results for the IBT - Demand for Approval Subscale. Dependency and care constructs will continue to be treated separately, awaiting a conjoint analysis in what follows. Subsequent care scale alpha-contribution plot analyses will be undertaken in conjunction with a conjoint principal components analysis allowing for the characteristics of each of the scales to be discussed in turn. Upon providing some preliminary evidence for the use of circular criteria in test development in the following section with the IBT - Demand for Approval Subscale, the section thereafter will focus on the development of the final Conventional Care Scales retained from the provisional pool of items. With a final set of scales in hand, subsequent and detailed analyses will follow. Scale Development and Fidelity to Circular Structure Criteria A theoretically informed approach involving substance plus structure is to be contrasted with a purely empirical or statistical approach (e.g., based on alpha and/or factor loadings for the decision whether items "hang together"). An empirical expression of this would be to demonstrate the circumplex's assistance in the selection of items maximally contributing to scale reliability, and thus, homogeneity. By extension of Circumplex Proposition 1, angular proximity within the circle is a direct measure of the similarity of constructs in interpersonal content, items located closer towards the scale's circular mean should provide the greatest alpha-contribution to the reliability of a scale as a whole. From Circumplex Proposition 2, this will be most true of items reflecting similar substantive or thematic content. Item alpha-contribution is calculated as the difference between the scale's alpha with and without the item. The larger the positive difference, the greater is an item's contribution to a scale's homogeneity. The current analyses are an extension of the only other study known of its kind where two interpersonal scales assessing loneliness and assertiveness, respectively, were examined (Gurtman, 1993) with confirming results: item proximity to the parent scale's circular mean is associated with its contribution to scale reliability. More generally, "the results demonstrate that the Interpersonal Circumplex is 70 more than simply a descriptive model, but can play a prescriptive role in a rational-theoretical approach to test construction" (Gurtman, 1993, p. 256). The circumplex's prescriptive role is thus two-fold. First, in its expression of what maturity entails, which has been discussed with respect to two criteria: behavioral flexibility in one's capacity to enact the complete circular array of interpersonal traits as the context demands, in contrast to excessive rigidity expressed in extreme high or low scores across octants (i.e., a central profile); and, optimal characterological style (i.e., moral traits), characterized as a blend of dominance and nurturance (i.e., Octant NO, Gregarious-Extraverted). The second prescriptive role for the circumplex is in its suggestion of criteria for item inclusion within parent scales. Following a variant of Gurtman's (1993) procedure, a best-fit quadratic trend line was drawn through 10 of the 12 points of the Irrational Beliefs Test - Demand for Approval Subscale (Jones, 1969) used in the current study (see Figure 11). As can be seen, the two items on the right, indicated by a hollow marker, were scale outliers; items 12 and 2 (identified in Appendix B) both contributed negatively to the scale's total alpha level (alpha-contribution = -.01 and -.004, respectively). Ideally all items should contribute uniformly to the parent scale's reliability and possess a narrow circular distribution. The reliability of the scale is taken as an index of its homogeneity and is calculated by an alpha-coefficient. The hypothesized expectancy within a dispersion of scale items is indicated by a shallow conic or inverted-U curve lying symmetrically about the circular mean. The circular mean is located at 0° on the X-axis of the alpha-contribution plot. As an item's angular projection deviates from the parent scale's circular mean, alpha-contribution decreases. Thus, scales with wider circular dispersions should evidence a deeper conic trend line. Note in Figure 11 that the apex for the Demand for Approval Scale would be located where its circular mean lies if it were recalculated less the two item outliers. A numeric index of this graphic point is derived by correlating item alpha-contribution with the (absolute) angular deviation from the scale's circular mean. Including all twelve items, alpha-contribution does decrease with proximity 71 Figure 11 Item Contribution to Alpha as a Function of Deviation from the Circular Mean for the IBT -Demand for Approval Subscale to the circular mean (r = -.69). Secondly, it is also hypothesized that, to the degree an item is interpersonal (i.e., projects into interpersonal space), as indicated by an item's vector length, an item should positively contribute to the total alpha. Thus, the current study proposes a second criterion of item desirability. From Circumplex Proposition 1, angular proximity within the circle is a direct measure of the similarity of constructs in interpersonal content, an item's (circular) factor loading can be determined from Equation 1 (duplicated below): V, * cos(0i - </>) (1) where <j>. is the angle (in radians) defining the midpoint of the item's octant category (i.e., 0°, 45°, 90°, etc.) and 0, is the angular projection of item i belonging to the relevant scale. An item's factor loading can be considered as a theoretically informed counter-part to item reliability. An item's factor loading combines the hypothesis regarding proximity to the 72 circular mean (Gurtman, 1993) with what should be expected of an item's vector length, that its vector length is a measure of interpersonal content and thus reflects its fit for an interpersonal scale. The correlation of the factor loadings of the twelve IBT - Demand for Approval items with angular proximity to the circular mean was strong (r = .55). Similar analyses will be undertaken on each of the conventional care scales as they are taken up in turn below. By way of brief review, item analyses involving the measure of dependency used in this study, items from available care scales, and items generated by test-writers for this study, mirror the findings in Pincus and Gurtman (1995) - both dependency and conventional care are interpersonally diverse constructs. Individual dependency and conventional care scales may be both drawing from an overly broad array of meanings, and by extension are also largely redundant to one another. If the current construct "sprawl" is to be curtailed and/or guided in some theoretically meaningful way, the acknowledged multifaceted nature of both conventional care and dependency constructs requires some structural underpinnings which a circumplex analysis can address. While the intent of the Pincus and Gurtman (1995) study was not to investigate item characteristics beyond the level of their circular projection, the alpha-contribution as well as circular characteristics of individual care/dependency scales will be conducted below. These analyses will be undertaken in conjunction with the development of additional scales employing the circumplex in its prescriptive role so as to more fully "map out" the domain of conventional care. Descriptive and prescriptive uses of the Interpersonal Circumplex can be of significant help to test developers. Integrating a circular structure approach within traditional substantive approaches to test development has not yet been done. A proposal for such criteria follows. Conventional Care Scale Development Gurtman's (1993) study, and the current research that follows it, suggests a potentially important role in the prescriptive use of the Interpersonal Circumplex for the development of interpersonal scales. Thus far, however, this has only been given some theoretical 73 articulation. What remains to be done involves the extraction of criteria for the purpose of test development which can be integrated into more traditional substantive approaches (Jackson, 1970). The current study proposes two criteria which follow from the above: (a) an angular proximity of ± 30° within some target angle (theoretically a priori or an a posteriori angular location marked by the construct), and (b) high item factor loadings determined by Equation 1 above. Projecting the 67 self-directed conventional care items generated by item-writers onto circular space reveals a broad array of interpersonal meanings spanning almost 270° (see Figure 10 above). (The other-directed and socially prescribed dimensions of conventional care will be discussed separately.) The development of theoretically determined scales according to both circular-theoretic and empirical methods proceeded as follows. Combined with traditional test item criteria, items were dropped if: • less than 7% of endorsements were located in three adjacent rating points on the 7-point scale (Livesley, Jackson, & Schroeder, 1989) (Highly skewed items, where less than 7% of endorsements were located in adjacent • rating points, were eliminated as they pose problems with floor and ceiling effects. This is especially problematic when their intended use includes both normal and clinical populations. In these circumstances scales will provide little information.) • standard deviations were less than 1 • angular projections were within range of their respective target (octant midpoint ± 30°) • corrected item-total correlations exceeded .40 on their respective scales • For each theoretically determined Self-directed Conventional Care Scale (i.e., Perfectionistic Care, Self-sacrificial Care, and Anxious Concern), final items were then composed of those eight items with the highest circular factor loadings on their respective octant midpoint (i.e., 270° and 315° for Octants JK and HI, respectively, see Equation 1, above). • A final principal components analysis with varimax rotation was conducted as an additional confirmation that items substantively "hung together." Items should load above .40 on their respective factor and not load above .40 on more than one factor. The resultant items should not only empirically "hang together" by the traditional substantive criteria, but the inclusion of circular structural criterion also necessitates a match based on a 74 "theoretical criterion for goodness" (Gurtman, 1993, p. 246). Loadings for the three jointly factored conventional care scales are reported in Table 4. While the initial principal components analysis indicated five factors with an eigenvalue greater than 1, the scree plot suggested a three-factor solution (Cattell, 1966). The repeated analysis signifying a three-factor solution accounted for 45.9% of the variance. In addition to the three self-directed Conventional Care Scales, two additional scales were constructed: Other-directed and Socially Prescribed Conventional Care Scales. As these were not expected to be interpersonally traited scales, but rather involve expectations and perceptions of others, respectively, the above circular criteria (proximity to a target projection and circular factor loading) were not required. The use of the circumplex in its descriptive role as a nomological net was still retained for all scales. In total five scales were developed: three self-directed scales (i.e., Perfectionistic Care, Self-sacrificial Care, and Anxious Concern), an Other-directed, and a Socially Prescribed Conventional Care Scale. Factor analyses for the two non-self-directed Conventional Care Scales are reported in Table 5 below. The initial principal components analysis indicated four factors with an eigenvalue greater than 1, while the scree plot suggested a clear two-factor solution accounting for 39.7% of the variance. Descriptive statistics for the Conventional Care Scales are also reported below in Table 6. Anticipating a more detailed discussion on gender below, apart from the Other-directed Conventional Care Scale, no gender differences among care/dependency scales were found. Separate means and standard deviations are reported for the Other-directed Conventional Care Scale. Questions of reliability and validity must be investigated before any instrument can be considered an adequate measure of the construct of interest. Most scales are designed to measure just one construct, to be homogeneous. The purely statistical criterion for "scaling" tests for homogeneity is to calculate a coefficient alpha (a). Alpha coefficients in the current study were computed as indicators of the internal consistency of items comprising each of the respective Conventional Care Scales and apart from the Socially Prescribed Scale all I. V. 75 Table 4 Factor Loadings for Self-directed Conventional Care Scales Factor Items grouped by scale 1 2 3 Anxious Concern 1. I get anxious at the possibility of saying something tactless with people .65 .15 .19 2. I worry if there is a flaw in how I have treated others .67 -.14 .11 3. I often worry that I will say something inconsiderate to friends .64 .29 .21 4. I fear that I will say something uncaring to friends .61 .19 .24 5. If I am not warm and caring towards others, I feel worthless .53 .26 -.04 6. In order to be a good person, I must always be gentle and nice .70 .13 .18 7. I feel completely selfish when I have to say "no" to friends .54 .12 .24 8. I feel totally selfish when I think about meeting my own needs .62 .17 .38 Self-sacrificial Care 1. The most important thing for me is to take care of the needs of others .23 .09 .78 2. Before I can be happy, others have to be cared for first .31 .28 .60 3. When I am able to make my friends happy (first), then I can feel happy .15 .25 .40 4. My happiness is not nearly as important as the happiness of those I care for .18 .14 .63 5. Putting other people's needs ahead of my own is not the most important thing (R) .21 .02 .56 6. Other people's needs are more important than my own .04 .09 .78 7. I am always ready to give others whatever they need .38 -.11 .47 8. I can't relax until I have done all I can for people that are important to me .18 .38 .48 Perfectionistic Care 1. It is important to me that I be perfectly caring towards others .07 .70 -.03 2. I primarily define myself as warm and kind .33 .57 .01 3. I must be compassionate and caring at all times .18 .78 .06 4. I demand nothing less from myself than complete sensitivity to others .16 .72 .11 5. I am very sensitive about not hurting the feelings of friends .12 .55 .32 6. One of the most important things to me is to avoid being unkind .31 .43 .11 7. One of the most important things is to avoid being selfish -.02 .59 .22 8. I expect nothing less from myself than perfect kindness to others -.07 .47 .39 Note. Items followed by (R) are reversed. Loadings > .40 are indicated in bold. 76 Table 5 Factor Loadings for the Other-directed and Socially Prescribed Conventional Care Scales Factor Items grouped by scale 1 2 Other-directed 1. Other people should be willing to interrupt their plans to make room for my needs .54 .21 2. If my friends weren't so selfish, they would help me more .64 .35 3. Those close to me should take more responsibility for the poor choices I have made .71 .06 4. I feel that my friends should solve my problems .66 .15 5. My friends should just realize that I have a great many needs .42 .39 6. I can accept that people need to take care of themselves and not just others (R) .60 .03 7. Those close to me are too busy to find out what I might need .60 .03 8. If my friends weren't so selfish I would be doing a lot better .63 .26 9. I can accept that friends won't always be warm and caring .48 .08 Socially Prescribed 1. It seems the more caring I become, the more that is expected of me .11 .62 2. People expect more care and concern from me than I am able to give .13 .62 3. I get stressed out by friends' expectations that I be completely responsive and caring .22 .59 4. Although they would never express it, my friends get upset when I fail to be completely warm and understanding .15 .54 5. People expect me to always be attentive to their needs .14 .67 6. Those close to me make me feel responsible to make things work out for them .35 .51 7. I feel overly burdened with responsibility -.08 .64 Note. Items followed by (R) are reversed. Loadings > .40 are indicated in bold. exceeded .80, desirable for scales of this size. Alpha coefficients are indicated along with other descriptive statistics in Table 6. 77 Table 6 Descriptive Statistics for the Conventional Care Scales Total Males Females Measure M SD a M SD M SD Self-directed Perfectionistic Care 37.7 7.04 .83 Self-Sacrificial Care 31.0 7.43 .81 Anxious Concern 31.1 8.02 .80 Other-directed 23.7 7.50 .80 25.1 7.54 22.9 7.3 Socially prescribed 26.2 6.37 .73 Note. Higher scores reflect greater levels of self-directed, other-directed, and socially prescribed conventional care. In and of itself alpha coefficients can not guarantee homogeneity; scale items may survive the screening process due to shared distress characteristics (Gurtman, 1993). Circumplex-based item statistics offer an additional theoretically informed criterion for item goodness. The correlation of any item with the central tendency of the construct in general can be computed by the circular factor loadings (see Equation 1 above), and were included in the criteria for item selection for the Conventional Care Scales. Those items with the highest circular factor loadings should, when combined, yield scales with optimal internal reliability. Across those self-directed scales developed outside the current study the correlation between circular factor loading and alpha-contribution ranged from .55 for the IBT - Demand for Approval Scale to .83 for the STSS - Silence Subscale, indicating their close relationship. The outlier was the STSS - Care Subscale (r = -.43); understandable given the heterogeneity and poor interpersonal characteristics of the scale generally. This point will become clearer as the psychometric characteristics of each scale is discussed in what follows. From an interpersonal perspective, then, questions of reliability are closely related to questions of validity where the content of the circumplex is used as a nomological net. An examination of 78 validity will occupy the bulk of this chapter's analyses and discussion. Another method of determining the validity of a measure is to attempt to reduplicate the scale's factor structure in a second sample. To test whether the factor structure is similar across two samples, a strict test of the factor structure's replicability was be conducted by computing a coefficient of congruence (Harman, 1976). As a measure of similarity between factors, the congruence coefficient is calculated on corresponding factors between two samples as the sum of cross-products of two sets of column-normalized factor loadings. In Study 2 (see above for details regarding sample), a strict test of the factor structure's replicability was conducted and a coefficient of congruence (Harman, 1976) was computed. The respective coefficients of congruence were .94 for the first factor (Perfectionistic Care), .92 for the second factor (Self-sacrificial Care), .93 for the third factor (Anxious Concern), .86 for the fourth factor (Other-directed Conventional Care), and .88 for the fifth factor (Socially Prescribed Conventional Care). The magnitude of these coefficients of congruence indicates a strong similarity in the factor structure across the two samples (Harman, 1976). Note that these coefficients, however, were calculated for scales somewhat larger and distinct from the final scales. Study 2 was conducted prior to the inclusion of circular structure criteria which resulted in a slightly smaller set of items. The items for the final scales reported in Study 1 represented 73% of the larger set of items used in Study 2. An integral component of Jackson's "minimum redundancy" method of test construction (e.g., 1974), for example, is the inclusion of a criterion that would reject items correlating too highly with social desirability, in addition to other non-relevant scales. Jackson's procedure is intended to produce scales with high content saturation, which while having some practical benefit where multiple measures are used to predict an external criterion, the implications for theory testing may be less than desirable. For example, it may be hypothesized that a particular scale be correlated to a certain degree with social desirability. While this criteria was rejected in the current study for reasons just stated, the question of whether a scale actually assesses its proposed thematic content, rather than 79 stylistic characteristics such as socially desirable responding, is an important one. Amongst the battery of measures a short form of the Marlowe-Crowne Scale (MC; Strahan & Gerbasi, 1972) was administered as an indicator of socially desirable responding. Of the self-directed care and dependency scales, only Perfectionistic Care and Unmitigated Communion evidenced a positive but low relation with the Marlowe-Crowne Scale (rs = .13 and .12, respectively, ps < .05). Among the non-self-directed scales, Other-directed Conventional Care and Social Perfectionism also correlated with the M C (rs = -.13 and -.22, ps < .05 and .001, respectively). The extent to which correlations with the M C scale reflect social desirability bias or substantive personality variance in the scale continues to be a point of disagreement (Paulhus, 1991). Nonetheless, these values are typical of many obtained in the present study between the M C and other measures, including (low) self-esteem (r = -.13), conscientiousness (r = .15), and neuroticism (r = -.12). In summary, together with a group of item-writers and guided by a close review of the literature, a total of 137 provisional test items was generated involving three dimensions of conventional care: self-directed, other-directed, and socially prescribed. The self-directed dimension was developed with several (sub)scales in mind with the intention of "mapping" the domain of conventional care onto circular space. The structural model employed in the current study was derived from the extraction of two principal components which were rotated to a theoretic set of circular criteria, enabling a description of the two extracted principal components as dominance and nurturance. Dependency items and then care items were projected onto circular space after a description of the process was detailed in the previous section. A circular projection of the self-directed conventional care and other care scale items (i.e., Silencing the Self Scale, Unmitigated Communion Scale, Social Perfectionism Scale - Self-directed) revealed a broad circular distribution similar to the study of Pincus and Gurtman (1995) examining the construct of dependency. The interpersonal circle was used prescriptively as an integral step in the construction of additional self-directed conventional care scales by including two additional criteria guided by the 80 circumplex proposition, angular proximity within the circle is a direct measure of the similarity of constructs in interpersonal content. An investigation of the Irrational Beliefs Test - Demand for Approval Subscale (Jones, 1969) used in this study confirmed and added support to an earlier study (Gurtman, 1993) reporting the use of circular criteria in predicting item alpha contribution. The farther an item was from the construct's central tendency in interpersonal space (i.e., the circular mean), the less it contributed to the parent scale's reliability. In what follows alpha-contribution plots for the individual care scales will be examined; as the IBT - Demand for Approval (dependency) Subscale was above. In addition to a continued investigation into the benefits of using circular criteria (i.e., Circumplex Proposition 1) for interpersonal test construction, Circumplex Proposition 2 stated that while sharing the same interpersonal structure, scales quite distinct from one another can occupy the same octant space. This has already been seen for the super circumplex itself. While respective octant scales for the IASR and IIP-C share interpersonal structure (i.e., are located within the same octant), they also possess variance that is unique to that particular "level" of construct. As was mentioned above, the problems form of interpersonal behavior (IIP-C) correlates only moderately high with their respective interpersonal traits (IASR). Similarly, scales occupying two adjacent octants could correlate highly with one another should they share a high component of Neuroticism, for example. The necessity of ipsatizing IIP-C octant scales by removing the general distress factor is one extension of this. The relevance for scale construction is that without circular structure analyses, items which are inappropriate according to their content can survived item screening owing to their shared levels of distress (e.g., neuroticism; Gurtman, 1993). According then to Circumplex Proposition 2, in addition to a circular analyses of individual parent scale items, a substantive approach will simultaneously be examined using a conjoint principal components analysis. An examination of the convergence of these two approaches was explored where it was hypothesized that a strong relationship would be found between an item's (a) contribution to 81 the parent scale's reliability and (b) its angle of projection relative to the central tendency of the parent scale. Scales which satisfy both (a) and (b) should also (c) uniformly load on their respective factor derived from the (substantive or content driven) factor analysis. In short, theoretically and empirically optimal scales should possess both adequate substance plus (circular) structure. The interpersonal circle will also be used in its more usual descriptive form, as a nomological net in which to examine questions of validity among the conventional care scales. The intention is to fill out a description of behaviors based on previous studies associated with the octant into which respective scales were projected. It further remains to be seen as to whether the various conventional care scales sample from the potential array of meanings in a broad overlapping fashion or whether they reflect independent and non-redundant facets of the conventional care construct. An examination of all self-directed care scales, including the Conventional Care Scales developed in this study, was undertaken for each of the relevant octants, which are primarily Octants (HI) Unassured-Submissive and (JK) Unassuming-Ingenuous. Preliminary Orientation to Scales and their Projections Before proceeding to the more detailed octant-by-octant investigation, it may be helpful to make some overall comments regarding the octant locations of the various scales and their intercorrelations. First, while redundant to information already present in the iconic representations above, the simultaneous plotting (or mapping, on the original metaphor) of circular means is indicated below in Figure 12. Recall the self-directed domain was to be undertaken first, prior to the investigation of other-directed and socially prescribed dimensions of conventional care. The construct of dependency will be included in the analyses that follow until greater clarity is gained regarding the similarities and distinctions between the two constructs. Second, the various measures of what has here been called conventional care, including the three scales developed in this study (i.e., Perfectionistic Care, Self-sacrificial 82 Figure 12 Projection of Self-directed Conventional Care Scales onto the Super Circumplex (DOM + ) (PA) Assured-Dominant (DE) (HI) Unassured-Submissive Note. 1 = IBT - Demand for Approval Scale (Jones, 1969) Silencing the Self Scale (Jack, 1991) 2 = Care 3 = Silence 4 = External 5 = Unmitigated Communion Scale (Helgeson & Fritz, 1998) 6 = Social Perfectionism Scale - Self-directed (Wiebe & McCabe, 1998) Conventional Care Scales 7 = Perfectionistic Care 8 = Self-sacrificial Care 9 = Anxious Concern Care, and Anxious Concern), were intercorrelated with one another. The diversity of interpersonal content, in addition to unique substantive variance, is reflected in the range of scale intercorrelations (rs = -.01 to .62; see Table 7). Other-directed and socially prescribed scales are included at this point for completeness. 83 Table 7 Intercorrelations among Care Scales 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12 1. Demand for Approval (IBT) .33 .29 .58 .36 .39 .29 .42 .52 .31 .65 .41 2. Perfectionistic Care (CCS) .58 .46 .00 .31 .42 .23 .24 .47 .37 .01 3. Self-Sacrificial Care (CCS) .47 .17 .45 .54 .24 .39 .60 .31 .07 4. Anxious Concern (CCS) .37 .47 .48 .51 .58 .38 .60 .35 5. Other-directed (CCS) .48 .19 .36 .40 .10 .36 .53 6. Socially Prescribed (CCS) .33 .33 .44 .39 .39 .45 7. Care (STSS) .35 .46 .41 .41 .16 8. Silence (STSS) .54 .12 .46 .31 9. External (STSS) .30 .55 .39 10. Unmitigated Communion (UCS) .27 .09 11. Social Perfectionism (SPS-SD) .49 12. Social Perfectionism (SPS-OD) Note. Correlation coefficients greater than .11 are significant at p < .05, those greater than .15 are significant at p < .01, and those greater than .18, p < .001. A cursory examination of zero-order correlation coefficients across the above self-directed conventional care and dependency scales is supportive of Circumplex Proposition 1 above: self-directed scales closer in proximity (e.g., within the same octant) are for the most part more highly correlated with themselves than with scales in adjacent octants. The average intercorrelation of the five scales within Octant HI (Unassured-Submissive) and the 84 four scales in Octant JK (Unassuming-Ingenuous) is .54 and .50, respectively. Correlations between scales across the two octants are for the most part low to moderate. For example, the correlations of Unmitigated Communion with and Silence and Social Perfectionism are low (rs = .12 and .27, respectively). Higher correlations across octants, such as between Care and External (r = .46), remain less than the mean intercorrelations among scales within octants. Two comments: first, individual scales with broader overlapping item distributions will inflate these across-octant intercorrelations, posing an interpretive confound. Secondly, according to Circumplex Proposition 1 we should expect that adjacent octants would moderately correlate, though less with octants further removed. Projected individual or group circumplex profiles, whether of the IAS-R or the IIP-C, typically express a peak in one of the octants with elevated but lesser levels in adjacent octants. The lowest values are in those octants opposite to elevated octants. Figure 13, for example, would be a typical individual or group profile, expressive of (JK) Unassuming-Ingenuous interpersonal functioning. Octant scores have been converted into T-score form (M = 50, SD = 10). The JK Octant is significantly elevated above the mean and significantly constricted in its opposite Octant B C and PA (Arrogant-Calculating and Assured-Dominant, respectively). (This is actually the reported IAS-R profile of the women in the current study - to be discussed below under the heading "Gender Differences: Factor Structure Measures.") A conjoint principal components analysis of all conventional care and dependency items is undertaken below to compare the substantive content of scale items. A discussion of the thematic homogeneity of individual scales will be conducted examining item alpha-contribution to scale coefficient alphas. Finally, the scales will be discussed in relation to the higher-order nomological net itself, the Interpersonal Circumplex - specifically the meaning of the octant into which the scale is associated via projection. The following analyses will be undertaken by octant, beginning with HI (Unassured-Submissive). 85 Figure 13 Typical Individual or Group Profile for the IASR or IIP-C Circumplex Data (DOM + ) (PA) Assured-Dominant 60 (HI) Unassured-Submissive Introduction to Conjoint Principal Components Analyses First, a conjoint principal components analysis was conducted across the nine self-directed conventional care and dependency scale items (73 items in total). In addition the ICQ - Negative Assertion Subscale was included to investigate the relation between negative assertion and self-silencing. Although the initial analysis indicated 21 factors with eigenvalues greater than 1, a scree plot suggested a five- to seven-factor solution. A six-factor solution was determined to be the most meaningful, accounting for 39.5% of the variance. Extracting a seventh factor contributed only an additional 2.2% variance. An examination of the factor loadings by octant is undertaken below. (See Appendix D for the complete table of factor loadings.) The relations between four item-indices were 86 investigated: (a) item factor loadings to examine the potential substantive item overlap among scales, (b) item circular projections, (c) item contribution to the reliability of parent scales using the alpha-contribution plot discussed above (Gurtman, 1993), and (d) face examination of item thematic content. As indicated in Figure 12 above, all nine scales, including the Demand for Approval Scale, fell as expected into Octants EQ and JK. Empirical investigations (i.e., mappings) will focus on these two octants (HI and JK) and will be proposed as the "two faces of conventional care." In the section that follows thereafter, some speculation as to potentially further conventional care constructs falling within Octants F G and NO will be undertaken. We begin with Octant EQ. Conjoint Principal Components Analyses: Octant HI (Unassured-Submissive) Four of the eight conventional care scales, in addition to the dependency scale, projected into Octant HI (Unassured-Submissive): Silencing the Self- External and Silence Subscales (Jack, 1991), Self-directed Social Perfectionism (Wiebe & McCabe, 1998), IBT -Demand for Approval Subscale (Jones, 1969), and one of the conventional care scales developed in the current study, CCS - Anxious Concern (see Figure 12). From the conjoint principal components analysis, Factors 1,3, and 6 best represent the substantive or thematic scale contents within the Octant HI; most highly represented by the Anxious Concern, Silence, and Demand for Approval (Sub)scales, respectively. Buhrmester et al.'s (1988) ICQ - Negative Assertion Scale was reverse scored and also included in the analysis to examine its relation with Silence. Negative Assertion loaded on a distinct fourth factor. Four substantive lower-order factors in total, then, were found to project within Octant HI. While sharing the same underlying interpersonal characteristics each possesses variance unique to itself (see Circumplex Proposition 2 above). Anxious Concern (CCS). Beginning with the Anxious Concern Scale, of its eight items all but three have their highest loading onto Factor 1. Two sample items, including their angular location, octant, and factor loading (see Equation 1 above) are listed below: 87 Angle Octant Loading Anxious Concern Items 256° FU .50 I get anxious at the possibility of saying something tactless with people 278° HI .47 I worry if there is a flaw in how I have treated others Social Perfectionism - Self-directed (SPS-SD). Of the 10 items within the Self-directed Social Perfectionism Scale, seven items had their highest loading onto Factor 1 with Anxious Concern. Five of these seven items projected into Octant HI with one just over the boundary into F G (239°). A l l of the items projecting into Octant HJ loaded greater than .40 onto Factor 1. Two sample items are given below. Angle Octant Loading Social Perfectionism Items 259° FU .64 I am afraid of making mistakes in conversations 263° FU .58 I am disappointed in myself when I say the wrong thing to somebody At the thematic level, Anxious Concern and the majority of Self-directed Social Perfectionism items share affectively negative concerns elicited within interpersonal contexts. What standards of valuation are involved when these individuals are so readily poised to make such self-statements is unknown. Are high scoring individuals actively pursuing high standards of social sensitivity, as the descriptor Social Perfectionism might seem to indicate, or is the social concern more negatively motivated, avoiding making social/conventional mistakes related to a need to assuage others' potential critical evaluations? Questions like the level of social skill high scorers possess, and to what extent these are merely perceived inadequacies, need to be addressed. In addition to this central cluster of socially anxious items from the Self-directed Social Perfectionism Subscale that loaded onto Factor 1, the remaining four items projected into Octants JK or L M . Examining the alpha-contribution plot confirms these four items likely belong to a JK scale or scales (see Figure 14). As the (absolute) angular deviation from the circular mean grows larger, an item's alpha-contribution to the overall Self-directed 88 Figure 14 Item Contribution to Alpha as a Function of Circumplex Angular Location for the Self-directed Social Perfectionism Scale .07 -r D -05 --.03 A , , , , 1 -90 -45 0 45 90 135 angle Social Perfectionism subscale goes down (r = -.67). (Note that alpha-contribution plots for the Self-directed Conventional Care Scales are not presented since it was precisely according to these criteria that scale items were selected.) Both the conjoint factor analysis and the alpha-contribution plot, in addition to an examination of the thematic content, suggests that the Self-directed Social Perfectionism subscale represents a large cluster of anxious items and a smaller heterogeneous set of additional items. Two items from the smaller cluster are reverse items which can be difficult to write. These often load upon their own separate factor, particularly when the negation "I don't..." is used rather than bipolar content. They can be difficult for respondents to interpret when disagreeing with the item, for example in the first two items below. Does disagreeing with striving for perfect relationships mean you are not "striving" but nevertheless relationships are important or does disagreeing mean that relationships are of no importance to you? The 89 third item below understandably loads onto Factor 5 with Perfectionistic Care (to be discussed below); it is thus the loading on Factor 5 which is reported below. The first two items indicated below failed to load onto Factor 1 with the majority of other Self-directed Social Perfectionism items. The three items from this second cluster set are: Angle Octant Loading Social Perfectionism Items 301° JK .12 It is not important to strive to have perfect relationships (R) 6° L M -.08 I don't expect my interactions with people to go perfectly (R) 308° JK .45 I make sure to do all the right things in relationships The italics in the above two items have been added to make the point that wherever a given thematic content (e.g., relationships, conversational interactions, etc.) is given agentic impetus, reflected in active qualifiers such as "striving," "making sure," "expecting," etc., such items will always project at higher elevations on the y- or dominance-axis. Individuals with IASR or IIP-C profiles projecting into the HI (Unassured-Submissive) octant may "hope" and "wish" for more satisfactory relationships, but are unlikely to actively pursue or maintain such relationships. The connection that apparently existed in the minds of the scale's developers was that motivational strivings towards perfect relationships was synonymous with anxiety about those relationships. But from the perspective of the Interpersonal Circumplex, meaningful differences can be detected at the level of a division into octants. It is this smaller second cluster of items within the JK (Unassuming-Ingenuous) octant that it would seem test developers (Wiebe & McCabe, 1998) had in mind when they entitled the scale "Social Perfectionism." That several of the 10 items failed to reflect the scale's specific intention signals the ever present possibility of discrepancy between an intended construct whose meaning is derived from some larger theoretical whole and the actual empirical pattern of a scale's correlates. Put elsewise, a (circular) structure clarification can contribute some theoretically important constraints on purely empirical or substantive approaches to test development. It 90 could be speculated that a "filling in" of the trend line of JK items with additional ones would place the circular mean of a perfection-in-relationships construct in conjunction with the perfectionistic care construct developed in this study. More will be said about the HI Octant and its nomological relations with other constructs in the following sections. External (STSS). The External Subscale possess a relatively narrow circular distribution of items (see Table 3 above: CM =251°, SCD = 14.9°) but with diverse thematic content. In addition to Anxious Concern and the larger subset of Self-directed Social Perfectionism items, most of the STSS - External (Jack, 1991) items projected into Octant HI. With respect to factor loadings, however, and similar to the Self-directed Social Perfectionism items, the 6-item External Subscale divides into two distinct clusters. The first three-item cluster loaded onto Factor 3 with Silence and fell in Octant HI; one of which cross-loaded onto Factor 2. The second two-item cluster loaded onto Factor 1, with Anxious Concern and the larger Social Perfectionism cluster, and fell either in, or on the border of, Octant FG. Angle Octant Loading External (self-critical cluster) Items 224° F G .60 I feel dissatisfied with myself because I am not able to do all the things students are supposed to be able to do these days 249° FG/HI .61 I never seem to measure up to the standards I set for myself Apart from the problematic reference to "students" in a broad scale intended for non-students, these two items possess a reflexively self-critical or introjective (Blatt, 1990) posture towards the self. The experience of a large discrepancy between the actual-self and external (ideal-self) standards results in depressive affect (Higgins, 1989). Two points need to be made here. First, while the External Self-perception Subscale is intended by Jack (in press) to measure "judging oneself by 'external' standards" (p. 14) there is, I think, at least two ways in which self-discrepancy can arise. Discrepancy can arise between the actual self and an ideal self, but it can also arise from a more horizontal comparison with simply an other-self. Regarding 91 the first form of discrepancy involving unrealistically high standards, there seems to already exist several good scales arising out of the perfectionism literature (Garner, Olmstead, & Polivy, 1983; Hewitt & Flett, 1991; Terry-Short, Owens, Slade, & Dewey, 1995). A second or horizontal form of discrepancy would seem, however, to be more closely aligned with a contextualized-self that takes its chameleon-like shape based on those in one's present company. The notion of a self that essentially lacks an integral self-identity is more germane to Gilligan's (1982) notions regarding loss of voice, dissociation, and voice-over. It is specifically with the absence of identity that one fails to "speak one's mind by telling all one's heart" (Rogers, 1990, cited in Gilligan, 1991, p. 21). The construct "external self-perception" may likely be a closely related construct to the third factor "other-directedness" (i.e., willingness to modify one's behavior to conform to other people) isolated from Snyder's Self-monitoring Scale (Snyder, 1974) by Briggs, Cheek, and Buss (1980). "Other-directedness" when understood in this context will hereafter be referred to as "other-focus" so as not to be confused with the "other-directed" dimension of the Conventional Care and Social Perfectionism Scales. In the first study to use the Interpersonal Circumplex to examine possible subcomponents within existing scales, Wiggins and Broughton (1985) found the other-focused scale component to project into Octant F G of the IASR. The second point regards the importance of distinguishing other-focus and/or external constructs from dependency related constructs such as compliance, suggestibility, need for advice and reassurance, and indecisiveness (Livesley, Jackson, & Schroeder, 1991). The only three items that appear to thematically capture the construct that Jack is seeking are: Angle Octant Loading External (true external cluster) Items 224° FU .40 I tend to judge myself by how I think my friends see me 269° HI .55 When I make decisions, my friends' thoughts and opinions influence me more than my own thoughts and opinions 257° HI .44/.41 I find it hard to know what I think and feel because I spend a lot of time thinking about how my friends are feeling 92 These three items possess an interpersonal thematic component that the above two External items do not. The third of the above three items also loads greater than .40 onto Factor 2 (reported to the right of the "/"), a factor more closely associated with Octant JK constructs. This last item above would appear to be a blend of "external perception" and sensitivity to how friends are feeling. Regarding Jack's (1991) External Self-perception Subscale, then, two subclusters are evident: a self-critical cluster and a true external construct cluster. I think the insight regarding the importance of a "external" construct is an important one. Jack's subscale, however, does a poor job of tapping it. Remaining is one item that loaded onto Octant JK Factor 2. Two items, then, the last one above and the one presented in the table immediately below, loaded onto Factor 2 along with Care, Self-sacrificial Care, and Unmitigated Communion to be discussed below. This diversity of content within the External Scale is reflected in the absence of an inverted-U alpha-contribution plot trend-line (see Figure 15). Angle Octant Loading External Item 266° FU .40 I often feel responsible for my friends' feelings Silence (STSS). So far we have discussed Octant HI scales Anxious Care, Self-directed Social Perfectionism, and External which all loaded in whole or part onto Factor 1, reflecting anxious concern within interpersonal contexts. The fourth scale projecting into Octant HI is the Silence Scale (Jack, 1991). A l l six of the items load strongly onto Factor 3, along with the three-item "external" subcluster reported above. Sample Silence items are given below: Angle Octant Loading Silence Items 245° HI .55 I don't speak my feelings to my friends when I know they will cause disagreement 252° HI .68 I try to bury my feelings when I think they will cause trouble with my friends 208° F G .49 When it looks as though the things I want in a friendship aren't there, I usually think that they weren't very important anyway 93 Figure 15 Item Contribution to Alpha as a Function of Circumplex Angular Location for the External Subscale .07 O O-.Ol --.03 -I 1 1 , 1 1 1 -45 -30 -15 0 15 30 45 angle A l l items but one project into Octant HI. The one F G item seems curiously out of thematic place. An examination of the alpha-contribution plot confirms this picture. The F G item appears as an outlier; contributing only negatively to the scale coefficient-alpha (alpha-contribution = -.015) and inflating the SCD reported in Table 3 above. The correlation between the Silence Subscale alpha-contribution and the absolute angular deviation from the circular mean is strong (r = -.91), even when the Octant F G outlier is removed from the correlation (r = -.46). With the FG item removed a predictable conic trend line is evidenced in the alpha-contribution plot (see Figure 16). This scale also reported the highest internal consistency of all conventional care scales developed outside this project (a = .83), even with the negatively contributing F G item. 94 Figure 16 Item Contribution to Alpha as a Function of Circumplex Angular Location for the Silence Subscale While the Silence subscale appears to have a number of good psychometric characteristics, the paucity of convergent and discriminant validation studies leaves it unclear what the scale is, in fact, measuring. Thompson writes "contrary to silencing the self theory, men, in this study, reported more global self-silencing [i.e., summing across the four of Jack's Silencing the Self Subscales] than did women" (1995, p. 348). She cites the research of Gottman (1994) who reports that it is men who typically resort to "stone walling." In contrast, women have been found to play an important role in maintaining relationships by introducing, elaborating, and confronting issues; men's role is a less expressive one. Gottman finds that men retreat from interactions using silent withdrawal, walking away, topic diversion, and generally minimizing their involvement in the issue. Walking out in the middle of an argument and slamming the door, does convey a message - simply not a direct 95 and verbal one. A central claim in Gilligan's and Jack's conceptions is the notion that women self-silence to preserve relationships; where the notion of "preserving relationships" is interpreted as something like "anxious attachment." In the face of evidence to the contrary Thompson seems to want to patch things up in a way that is contrary to the theory. It is women who are confrontive for the sake of preserving the relationship. Part of the problem is the conflation of two domains. Gottman's (1994) research is within the context of close and intimate relationships whereas Gilligan's work largely involves the domain of "friends." Harter, Waters, Whitesell, and Kastelic (1998) speak of "level of voice" and have shown that not speaking of opinions (i.e., silencing) occurs more within the context of the broader concentric circles of peer group relations and adults than it does closer, safer relationships. The footnote here is not intended to resolve the issue, but merely to note the importance of not mixing contexts. The fact remains that women do not silence themselves among "friends" more than do men and, while no gender difference was found in the present study, there is some evidence to suggest the reverse may be true. This raises the validation question. It could be questioned whether "not rocking the boat to avoid confrontation" is an accurate operationalization of Gilligan's notion of "silence" and "voice." It is important to note here some potentially significant distinctions between Gilligan's notion of "voice/silence" and related constructs such as "(reluctant) self-disclosure," "negative self-assertion," and "restricted expression of angry affects/anger-in." In an attempt to sharpen the contours of the notion of voice, Harter suggests that, "the voice construct refers to the expression of one's opinions or what one is thinking rather than to intimate information or expression of needs and wants" (Harter et al., 1998, pp. 893-894). The polar opposite of "voice" is "the suppression of one's true self" (Harter et al., 1998, p. 894, italics added), distinguishable from negative self-assertion. The absence of voice is what the Silence scale is attempting to assess. Anger-in. Regarding the first distinction between voice/silence and anger-in, 96 Spielberger, Johnson, Russell, Crane, Jacobs, and Worder's (1985) Anger-in Scale measures the frequency with which angry feelings are held in or suppressed (e.g., "Irritated more than people are aware"). Thompson brings the two together in the suggestion that, "research should examine the association between women's self-reported silencing the self and observed expression of anger and disagreement during conflict resolution" (1995, p. 350). In a preliminary effort to test the relation between anger-in and Silence, five anger-in items were identified within the provisional pool of conventional care items by their thematic content. An additional conjoint principal components analysis was conducted on items reflecting a proxy Anger-in Scale. The (reverse-coded) ICQ - Negative Assertion Scale was also included to complete the range of available comparisons. The initial analysis indicated three eigenvalues greater than 1 and the scree plot suggested the same. The analysis was repeated for a three-factor solution which accounted for 54.5% of the variance. Silence, (reverse-coded) Negative Assertion, and Anger-in items all loaded greater than .40 on their respective factors, and apart from one item in each scale loading above .40 on more than one scale, the three scales formed unique constructs. The complete table of item loadings is reported in Appendix E. The five items used as a proxy for anger-in are listed below with their angular locations, octant placement, and factor loadings. Angle Octant Loading "Anger-in" Items 242° HI .77 It's OK if I have disagreements with friends (R) 266° HI .45 Being a caring person means never getting angry 238° F G .42 Expressing anger towards someone close to me does not mean I am uncaring (R) 275° HI .76 I can accept that there will be times when I will come across as unkind or rude (R) 230° F G .65 I can still regard myself as a caring person even if I get into an argument with someone close to me (R) The circular dispersion of items would suggest anger-in's placement within FU (Unassured-Submissive) towards the side of F G (Aloof-Introverted). While located in the 97 same octant and thus sharing some underlying interpersonal characteristics as that of Silence, Anger-in appears to be a distinct construct, insofar as the current proxy scale is able to anticipate a more rigorous examination. Self-disclosure. In addition to distinguishing Silence from Anger-in, two further distinctions need to be made with silence: self-disclosure and self-assertion. Pearson and colleagues define self-disclosure as "intentional sharing of intimate information about oneself" whereas self-assertion is defined as "communication of one's needs and wants in an effort to gain another's volitional understanding and/or compliance" (Pearson, Turner, & Todd-Mancillas, 1991, p. 162). Self-disclosure and the self-assertion have consistently been identified as important interpersonal skills (Buhrmester et al., 1988) and their relation to the various conventional care scales will be examined below. Anticipating subsequent analyses, Silence and Self-disclosure were not found as opposite poles of the same continuum, where the latter is understood as the capacity to disclose personal information. It is important to note here that scales in circumplex space are all unipolar - each can be considered a negative item with respect to items in the opposite octant. Self-disclosure items do not project into PA (Assured-Dominant) as reverse items to Octant HI. Rather, the Self-disclosure items project narrowly into Octant NO (Gregarious-Extraverted: CM = 39°, SCD = 7.1°). The "problems" form of self-disclosure is captured in the following IIP-C items: "I tell personal things to other people too much" and "I open up to people too much." The intense form of self-disclosure involves offering excess information or types of information about oneself when it is not appropriate or requested. Greater elaboration on this construct will be undertaken in a following section entitled "Interpersonal Competencies." Negative Assertion (ICQ). Regarding the third distinction, between silence and (negative) self-assertion, the present study included Buhrmester et al.'s ICQ - Negative Assertion Scale, defined as involving competencies in the "assertion of personal rights and displeasure with others" (1988, p. 992). While Self-disclosure did not project into Octant HI 98 as reverse items, the Negative Assertion Scale items did. The (reverse-coded) scale projected narrowly into Octant HI (CM = 87°, SCD = 63°). It is a reasonable hypothesis that the construct of negative assertion is simply the reverse form of self-silence. This hypothesis was tested by including the Negative Assertion Scale in the conjoint principal components analysis. The results showed, however, that all eight items loaded strongly onto their own respective factor. Two of the total items with the highest loadings, expressing hurt and anger, respectively, are provided below. A more elaborate discussion of these points will be undertaken later. These results suggest that the voice/silence construct can be both theoretically as well as empirically distinguished from self-disclosure and negative assertion. Angle Octant Loading Negative Assertion Items 81° PA .75 Telling a companion that he or she has done something to hurt your feelings 89° PA .72 Telling a date/acquaintance that he or she has done something that made you angry Recall that, in the larger principal components analysis, three of the External items also loaded onto Factor 3 with Silence, and should therefore be understood as providing some additional information regarding what the Silence Scale is tapping. Persons who score high on the scale are simultaneously reporting that they defer to other people's opinions in making decisions (item 9) and in matters of self-understanding (item 3), and dissociatively "find it hard to know what I [they] think and feel" (item 15). The factor on which these items load would appear to be a good representative of Gilligan's concerns and has been here labelled a "False-self" factor. Demand for Approval (IBT). The fifth scale that projected into Octant HI, and the only dependency measure used in this study, was the IBT - Demand for Approval Subscale (Jones, 1969). Already noted above, most of the items projected into the Octant HI. The two items contributing only negatively to the scale's reliability were items from Octants JK and L M . (For further discussion regarding the alpha-contribution plot of the IBT - Demand for 99 Approval Subscale, see Figure 11 above.) Examining the conjoint principal components analysis reveals that seven of the twelve items had their highest loading on a distinct sixth factor; six of which were greater than .40. A subcluster of three items loaded onto Factor 1, along with Anxious Concern, the larger subcluster of Social Perfectionism, and the self-critical smaller cluster of External. These three items share with the other scales an anxious concern within interpersonal relations. In the Anxious Concern and Social Perfectionism Scales, however, worry/concern is reported to revolve around being tactless, inconsiderate, or otherwise uncaring within social exchanges. Concern in the dependency scale involves worrying whether others are accepting or rejecting. The close relation between the two speaks to the question of motivation. Individuals are anxiously concerned not to seek perfect relationships, but to avoid rejection. The self-critical subcluster of items from the External Scale is relevant here, relating particularly to the last item below. In comparison to the main cluster of the IBT -Demand for Approval items, this subcluster of three items falls more closely toward Octant F G (Aloof-Introverted) where low self-esteem and depression are projected (more will be said of this later). Loadings reported below are for Factor 1. Angle Octant Loading Demand for Approval Items 251° HI .52 I often worry about how much people approve of and accept me 250° HI .54 I have considerable concern with what people are feeling about me 231° F G .60 Noticing one fault of mine makes me think more and more about my other faults The bulk of the Demand for Approval items, however, do reflect the intended thematic content and substantively "hang together" on the same factor, Factor 6. The F G outlier, indicated in Figure 11 above, failed to load on any factor and its projection into Octant L M confirms it as a poor item. Three sample items from the main cluster are reported below. 100 Angle Octant Loading Demand for Approval Items 284° HI .45 It is important that others approve of me Although I like approval, it's not a real need for me (R) 286° HI .66 273° HI .57 If others don't like me that's their problem not mine By way of an octant summary, a principal components analysis of the 61 self-directed items, plus the ICQ - Negative Assertion and IBT - Demand for Approval Scales, resulted in the extraction of six factors. While the factor loadings reflect the substantive or thematic elements, their octant circular projections provide information regarding their superordinate (or underlying) shared interpersonal structure. Three of the principal components, or factors, are loaded by scales projecting into Octant HI. The first Octant HI factor, Factor 1, was reflected largely in three scales. The Anxious Concern Scale and the larger Social Perfectionism subcluster possessed an anxious interpersonal concern. Two items from the substantively heterogeneous External Scale and three items from the Demand for Approval Scale all shared an anxious concern regarding measuring up to standards and worry or concern regarding others' approval. Individuals whose peak profile is in Octant FU possess a sense of self that is determined by how others respond to them. It is for this reason that the Anxious Concern Scale was labelled "Concern" rather than "Care," leaving open the question of whom the object of concern was directed towards: genuine caring concern for the other or concern with the other in so far as it tells on one's own self-acceptance. A second Octant HI factor, Factor 3, was constituted primarily by the Silence Scale and three "other-focused" items from the External Scale. Finally, a third factor, Factor 6, was singularly occupied by the majority of Demand for Approval items. In addition to a principal components analysis and a circular structure analysis, alpha-contribution plots were investigated. Findings are supportive of the notion that an circular item analysis could be used as a prescriptive tool for test developers. Where a heterogeneity of substantive and/or thematic content was detected among items within a scale, it was found that item alpha-contribution decreased with their absolute distance from the parent scale's 101 circular mean. Alpha-plottings confirmed the findings of thematic heterogeneity among parent scale items. While circular criteria alone can not be used to determine scale-item fit, given Circumplex Proposition 2, they can indicate when items are not a good fit, by Circumplex Proposition 1. Conjoint Principal Components Analysis: Octant J K (Unassuming-Ingenuous) It was hypothesized above that examining the relation among conventional care scales within interpersonal space would provide some structural grounds upon which to define conventional care as multifaceted. Given that meaningful differences exist in interpersonal space, at least at the level of the octant, facets located within distinct octants should possess distinct characteristics relative to one another. By extension these same criteria were employed to create scales that were reflective of their respective octants (i.e., not excessively broad). It was expected that "two faces" of conventional care were to be found, Octant HI and JK, in that they are located on the caring side of the Nurturance axis and in the lower submissive half of the Dominance (or power) axis. Respective scales and factors for Octant HI have been discussed above, and will be followed below by an examination of Octant JK where a similar set of analyses will be undertaken. The four remaining self-directed scales projected into Octant JK (Unassuming-Ingenuous): the Care Scale (STSS), the Unmitigated Communion Scale, and two scales developed in the current study, Perfectionistic Care and Self-sacrificial Care (see Table 4 above). These four scales loaded principally onto two factors, and given that these two factors were most strongly represented by the two scales developed in the current study were entitled "perfectionistic care" and "self-sacrificial care." Perfectionistic Care (CCS). The first scale, Perfectionistic Care, reflects the thematic content of striving towards, or placing of first importance on, the care of others. A l l eight items loaded greater than .40 on Factor 5. (Again recall that a complete table of the conjoint factor loadings can be obtained in Appendix D.) Two sample items are: 102 Angle Octant Loading Perfectionistic Care Items 333° JK .61 It is important to me that I be perfectly caring towards others 318° JK .59 I must be compassionate and caring at all times None of the other self-directed care or dependency items loaded onto this fifth factor, indicating the substantive distinctiveness of the perfectionistic care construct - but for the one JK item from the Social Perfectionism Scale mentioned above. An additional item from the Social Perfectionism Scale loaded .39, both of which shared a more agentic care involvement (versus "worry about" interactions; see below). These two items would seem to capture better the label of "social perfectionism" that the authors had in mind. The greater motivational component within Octant JK is readily discernable and accounts for the higher projection along the dominance axis. That JK items noted above from Self-directed Social Perfectionism loaded onto the perfectionistic care factor would seem to indicate that striving to be perfectly caring also involves the desire to have good relationships. Angle Octant Loading Perfectionistic Care Items 308° JK .45 I make sure to do all the right things in relationships 323° JK .39 I have a strong desire for conversations to go well Self-sacrificial Care (CCS). The second scale projecting into Octant JK, Self-sacrificial Care, reflects the thematic content of putting others' needs and happiness before one's own. A l l eight items loaded greater than .40 onto Factor 2. Two sample items are: Angle Octant Loading Self-sacrificial Care Items 327° JK .64 Before I can be happy, others have to be cared for first 320° JK .65 Other people's needs are more important than my own Regarding the second factor expressive of the JK Octant, Factor 2, while all six items of the Care Scale had their highest loading here, only one exceeded .40. The circular dispersion of Care items reflects two subclusters, a JK subcluster and a smaller HI subcluster. The two JK items with the highest factor loadings are: 103 Angle Octant Loading Care Items 335° JK .51 Caring means putting other person's needs in front of my own 319° JK .38 In my friendships, my responsibility is to make the other person happy The two HI Care items below share Octant Ed's submissive and self-devaluing aspects: Angle Octant Loading Care Items 265° HI .32/.28 Caring means choosing to do what my friends want even when I want to do something different 272° HI .36A25 Doing things just for myself is selfish Both of these items have elevated secondary loadings on HI Factor 3 marked by the Silence scale, thus representing blends of Octant JK Self-sacrificial care and Octant FU submissiveness. (Factor loadings to the right of 7" indicate the secondary loadings on this second factor.) From the alpha-contribution plot for Care (see Figure 17), the two HI items can be seen on the left. The Care Scale possessed a number of undesirable psychometric characteristics. Shown in Figure 17 these two distinct clusters, one to the left (HI) and the other to the right (JK), resulted a U-shaped rather than an inverted-U trend-line. The correlation between angular deviation from the circular mean and item alpha-contribution was approximately zero (r = .05). The Care Scale evidenced the lowest coefficient alpha from among the self-directed scales (a = .64). In addition to low internal reliability, several of the items possessed low vector lengths, suggesting these items are low in interpersonal content: they likely reflect beliefs whose interpersonal implications can be varied if not unrelated to actual self-reported behavior. Vector lengths less than .15 can be considered low (cf. Alden et al., 1990). Vector lengths, and by extension circular item loadings, are by far the lowest for Jack's (1991) Care Subscale in comparison to the other subscales. The point will be made more graphically in subsequent analyses. (The mean vector lengths, and the related circular item loading, for 104 Figure 17 Item Contribution to Alpha as a Function of Circumplex Angular Location for the Care Subscale each of the self-directed scales was reported in Tables 2 and 3 above.) For example, the expression "Caring means ..." (items 1 and 7, vector lengths = .12 and .11, respectively) will likely have a more distant relation to self-reported behavior than the motivational "I strive to be Thompson (1995) also reports the Care Subscale possessing the lowest alpha coefficients among Jack's (1991) scales, although her findings are higher than the present sample (a - .74). Unmitigated Communion (UCS). Thus far, then, Perfectionistic Care loaded onto a distinct Octant JK Factor 5, and Self-sacrificial Care and (less strongly) Care Subscale loaded onto Octant JK Factor 2. In addition, the majority of items from the Unmitigated Communion Scale also either loaded greater than .40 or had their highest loadings onto Factor 2. Examining the distribution of Unmitigated Communion items reveals two items projecting into Octant L M and an additional far-flung item at 12 o'clock in Octant PA. 105 Examining the alpha-contribution plot (see Figure 18) indicates that the latter PA item (item 4) made one of the smallest contributions to the parent scale's coefficient alpha (alpha-contribution = .0138), second only to the negative contribution of the Octant L M item (item 5: alpha-contribution = -.0073). The second L M item (item 2) made the next least contribution to total alpha (alpha-contribution = .0148). The alpha-contribution plot indicates that these three items, possessing the greatest angular deviation from the circular mean, contribute either minimally or negatively to the scale as a whole. Apart from the Octant PA outlier and the small cluster of Octant L M items - which contribute little to the reliability of the scale and tap what will later be described as an Intrusive Care Scale - the majority of content within the Unmitigated Communion Scale reflects Octant JK Factor 2 themes of self-sacrificial care, where other's needs are deemed of greater concern than one's own. While the circular mean of the Unmitigated Communion Scale sits on the Octant JK-L M boundary (338°; see Figure 12 above), the removal of these three extraneous items would place the scale well within the JK Octant. Two representative items from the larger JK cluster are: Angle Octant Loading Unmitigated Communion Item 335° JK .63 I always place the needs of others above my own 322° JK .49 For me to be happy, I need others to be happy In summary, projecting into Octant JK were four self-directed scales: Care, Unmitigated Communion, and two scales developed in the current study, Perfectionistic Care and Self-sacrificial Care. These four scales were represented by two substantive factors, Factors 2 and 5. Loading uniquely onto Factor 5 was the Perfectionistic Care Scale. The other three scales loaded for their greater part onto Factor 2. The best representative of this factor was the Self-sacrificial Care Scale followed by the larger cluster within the Unmitigated Communion Scale. The Care scale loaded uniformly, though weakly, onto Factor 2 - perhaps owing to its low interpersonal item content. 106 Figure 18 Item Contribution to Alpha as a Function of Circumplex Angular Location for the Unmitigated Communion Scale •07 1 -.03 -I 1 1 , 1 1 1 -90 -45 0 45 90 135 180 angle In broad summary of scales projecting into Octants HI and JK, not only does the development of scales using the circumplex help greatly to clarify the structure-theoretical and thematic content of scales but also has important prescriptive implications towards developing homogeneous scales with such qualities as brevity where each item contributes proportionately to the overall internal reliability of the intended scale. The notion of an alpha-contribution plot (Gurtman, 1993) has proven valuable in the assessment of individual items in their relation to their parent scale. Graphically, the relation between item alpha-contribution and divergence from the scale's circular mean in a scale with a moderate spread of items is a conic or inverted-U trend line symmetric about the circular mean. The correlation between an item's alpha-contribution and angular deviation from the scale's circular mean proved another way of stating the relationship: items further from the circular mean contributed less, even negatively, to the overall scale's coefficient alpha. Based on 107 these results and those of Gurtman (1993), two circular criteria were included - in addition to the more usual criteria (i.e., item-total correlations, factor loadings) for item selection - in the development of the Self-directed Conventional Care Scales. The two circular criteria were: (a) item angular location within ± 30° of the target octant midpoint and (b) high circular factor loadings (see Equation 1 above). To facilitate a comparison of all self-directed care and dependency scales a principal components analysis was undertaken extracting six factors, including the ICQ - Negative Assertion Scale. In those cases where factors made distinctions within existing scales, these clusters were reflective of their thematic content and location in the alpha-contribution plot. Three of the factors fell within the HJ (Unassured-Submissive) octant and two fell with the JK (Unassuming-Ingenuous) octant. Omitting some of the smaller clusters discussed above, the six factors, factor labels, and their representative scales are presented below in Table 8. While the two octants, FU and JK, are largely representative of the two faces of conventional care, consistent with Circumplex Proposition 2, more than one factor and their representative scales fell within one and the same octant. Four factors fell within Octant FJI: Factor 1 -Anxious Self-critical Concern, Factor 3 - False-self, Factor 4 - Negative Assertion; and Factor 6 - Demand for Approval. Two factors fell within Octant JK: Factor 2 - Conventions of "Goodness" and Factor 5 - Perfectionistic Care. From among the care/dependency items, those scales developed in the current study loaded most uniformly on their respective factors (with the exception of the Silence Scale). From the above it can be seen that, apart from the Silence and Negative Assertion Scales, the other scales developed outside the present study either are not represented by very homogeneous scale items (External, Self-directed Social Perfectionism, to a lesser extent Unmitigated Communion and Demand for Approval) or they are redundant to and better represented by scales developed in the present study (Unmitigated Communion); they have poor psychometric characteristics on other grounds (Care); or do not map on very well onto the constructs to which they were intended (Self-directed Social Perfectionism and External). 108 Table 8 Summary of the Conjoint Principal Components Analysis Octant Factor Label Scale HI H I 1. Anxious Self-critical Concern 3. False-self HI(R) 4. Negative Assertion HI 6. Demand for Approval Anxious Concern Scale Self-directed Social Perfectionism: larger Octant HI cluster two self-critical items from External Silence three "other-focused" External items JK JK 2. Conventions of "Goodness" 5. Perfectionistic Care • Self-sacrificial Care • Unmitigated Communion • Care Note. HI(R) = Negative Assertion Scale reverse scored. For example, Factor 1, falling with Octant HI, was loaded by the larger Octant HI cluster of the Self-directed Social Perfectionism Scale, along with two self-critical items from the External Scale and the Anxious Concern Scale. Two further care constructs bear mentioning, the one locatable in Octant F G (Aloof-Introverted) and the other in NO (Gregarious-Extraverted). While an insufficient number of items were represented from which to construct representative scales, some more speculative discussion is warranted in the interest of "mapping" the conventional care domain. Continuing around the circumplex counter-clockwise, discussion of Octant NO will be undertaken next. 109 Further Mapping: Octant N O (Gregarious-Extraverted) A posteriori examination of item thematic content from among the conventional care provisional item pool by octant projection revealed five distinct items within Octant NO, composing a rudimentary Intrusive Care Scale. The five items are: Angle Octant Intrusive Care Items 19° L M / N O I must not allow my own problems to prevent me from helping friends whenever they need it 39° NO I do not get overly involved in my friends' problems (R) 43° NO Even though they are my friends, it is important that I not get overly involved in helping (R) 63° NO I feel distressed if friends do not tell me about their problems 32° NO I get overly involved in my friends' problems What is immediately noticeable in the above items is the increase in actual agentic involvement in one's acts of caring. That these items are located a good distance from scales projecting into JK and HI, in particular, where these items are virtually their reverse counterpart, should raise a certain skepticism regarding the actual activities of caring in which individuals reporting high scores of HI and JK are engaged. (Recall that characteristic profile configurations for individuals can be categorized by the peak octant elevation - see Figure 13 above.) Hypothesizing about the construct of dependency, Birtchnell (1987) posited the existence of a "possessive/caregiving/altruism" interpersonal style in Octant NO (see Figure 19). Drawing on the notion of "projective identification" (Kernberg, 1976), Birtchnell comments: He [sic] denies his own deficiencies and needs and projects them onto the other. His selfless and altruistic attitude makes it hard for others to be caring towards him. The devotion and gratitude of others may be the closest he gets to affection.... [He] comes to identify with the object of his directiveness and tends to live vicariously through him. So much attention is paid to the other and so much emotion is invested in him that attention is diverted away from the directive person whose own identity tends to become subsidiary. It is as though he inhabits the other person whose successes or failures become his. (p. 23) 110 Figure 19 Characteristic Attitudes and Behaviors in Circular Space Domination Suppression Intimidation Aloofness Destructiveness Seeking distance Seeking privacy Emotionally cold Secretive D E T A C H M E N T m Fear of involvement Fear of closeness Entitlement Narcissism Non-human attachment Addictions Fear of losing control Fear of being taken over Need to take the lead Need to be in control Seeking responsibility D l R E C T I V E N E S S RE C E P T I V E N E S S Seeking guidance Seeking reassurance Needing to follow Restrictiveness Possessiveness Care-giving Altruism Responsive to the needs of others H A T T A C H M E N T Care-eliciting Succorance Helplessness Seeking proximity Seeking company Emotionally warm Open Fear of losing contact Fear of being alone Fear of rejection Fear of assuming responsibility Note. Adapted from "Attachment-detachment, Directiveness-receptiveness: A System for Classifying Interpersonal Attitudes and Behaviour," by J. Birtchnell, 1987, British Journal of Medical Psychology, 60, p. 25. Copyright 1987 by the British Psychological Society. I l l In addition to the confusion Birtchnell creates by referring to "the directive person" rather than as a blend of directiveness and attachment as his figure illustrates, the "external self-perception" or chameleon-self that is mentioned might seem more fittingly to belong to Octant HI. Yet, in positing some theoretical hypotheses regarding extreme positions in the NO octant, he highlights a construct that has not been tapped by any measure apart from a couple of items in the IIP-C marking Octant NO: "It is hard for me to.stay out of other people's business" and "I feel too responsible for solving other people's problems." Bowlby (1980/1991) elaborated on the notion of "compulsive care-giving" from the perspective of attachment theory. Bowlby also provided anecdotal accounts of individuals who have developed a pattern of chronic mourning after the loss of someone, usually a spouse or child, who was the prior object of compulsive care-giving. Other individuals responded to loss by concerning themselves intensely and to an excessive degree with the welfare of others. Instead of experiencing sadness and welcoming support for themselves, they proclaim that it is someone else who is in distress and in need of the care which they then insist on bestowing, (p. 206) Elsewhere Bowlby states: The care they bestow may amount to an obsession; and it is given whether it is welcomed, which it may be, or not. It is given, also, whether the cared-for person has suffered a real loss of some kind or is only believed to have done so. At its best this caring for another person may be of value to the cared-for.... At its worst, it may result in an intensely possessive relationship.... In addition, the compulsive caregiver may become jealous of the easy time the cared-for is thought to be having. (1980/1991, p. 156) Bowlby rejects the object-relations notion of projective-identification but affirms the clinical description, "a compulsive caregiver seems to be attributing to the cared-for all the sadness and neediness that he [sic] is unable or unwilling to recognize in himself, the cared-for person can be regarded as standing vicariously for the one giving the care" (pp. 156-157). West and Sheldon-Keller (1994) have sought to operationalize Bowlby's attachment constructs for use with adults. An examination of the thematic content of scale items looks 112 very much closer to items projecting into Octant JK, than the more agentic NO items. Two items from their scale read: • I put my attachment figure's needs before my own • I don't sacrifice my own needs for the benefit of my attachment figure (R) A few items sound more like L M (Warm-Agreeable) items: • I enjoy taking care of my attachment figure On the other hand, reference to an "attachment figure" might locate the construct more within Octant L M . Why caring within close, especially intimate, relationships might be an uniquely Octant L M construct will be discussed below. Central to interpersonal theory (and family systems theory; Minuchin, 1985) is the notion of circular or bidirectional causality (Benjamin, 1974). Rather than viewing an individual's behavior as a direct result of situational or interpersonal motives, cause-to-effect, the focus is on a two-person mutual influence. Events that happen to us are in large part brought about by our own actions. Leary (1957) first asserted what would later become the principle of complementarity, that "the purpose of interpersonal behavior is to induce from the other person behavior that is complementary to the behavior proffered" (p. 112). About a decade later Carson (1969) gave an explicit definition, stating that complementarity occurs on the basis of reciprocity on the Dominance-Submission axis, and on the basis of correspondence on the Nurturance-Detachment axis (e.g., dominant behavior elicits submissive responses, while loving behaviors elicits loving responses). The principle of complementarity suggests that domination requires individuals who are willing to allow, or at least perceive it in their best interest to allow, domination; compulsive care-givers require dependent individuals who are willing to be cared for. Each position confirms and reifies the other's model of interpersonal relations: "I perceive myself as helpless and others as powerful and capable." Caregiving is the reciprocal of care-eliciting.... It may be easier for the carer to suppress his [sic] own dependency needs by emphasizing such needs in 113 another.... the satisfaction derived from the caring may be so great that the carer may find it hard to allow the other to progress.... The two most common defences of the directive individual are denial and projection. (Birtchnell, 1987, pp. 24-25, 23) By these lights, a picture of conventional care is incomplete without an understanding of compulsive care-giving, or what has here been called intrusive care, its complement. Further Mapping: Octant FG (Aloof-Introverted) Two items from the initial provisional pool generated by item writers projected into Octant F G sharing a moral language of "selfish" (the first two of the following items). Angle Octant Scale Selfishness Items 236° F G provisional To look after my own needs is completely selfish 239° F G provisional Taking better care of myself would be selfish 255° FG/Ffl Anxious I feel totally selfish when I think about meeting my own needs Concern 333° JK Care One of the worst things I can do is be selfish 272° FU Care Doing things just for myself is selfish In addition, the final item in the Anxious Concern Scale projected onto the FG/FJJ border (by the ± 30° criteria) and could be regarded as a peripheral item from a more extensive cluster located more centrally in FG. That the item also failed to load together with the other Anxious Concern scale items onto Factor 1 in the conjoint principal components analysis may also be an indication of its non-optimal placement. Two scales from the Jack's (1991) Care Subscale also contained a similar thematic content, but their low vector length may account for their deviation from their speculated F G placement. (Geometrically, small deviations in the correlations with dominance and nurturance will cause larger variability in angular projection among items with a short vector length.) Anticipating a series of hierarchical regressions predicting depression, some comment should be made regarding the prevalence of negative affect in Octant FG. The projected circular distribution of individual scale items of Beck's Depression Inventory (Beck, 1967), Rosenberg's Self-esteem Scale (Rosenberg, 1965) scored in the direction of low self-esteem, 114 and IASR-B5 Neuroticism (Wiggins, 1995) reveal scales with nicely constrained items all narrowly located in Octant FG. By Circumplex Proposition 1 it can be anticipated that care/dependency constructs located in Octant FG would be the most predictive of depression and negative affect. Depression has become perhaps the most investigated and discussed of (gendered) diagnostic categories, with twice as many women experiencing the disorder than men. This is true across white, Hispanic, and black ethnic groups (Nolen-Hoeksema, 1990). Jack's (1991) intention in developing her Silencing the Self Scale was to apply Gilligan's theory and methodology to understanding women and depression. Jack (in press) writes: My work on silencing the self theory began by listening.... closely to the moral themes in depressed women's narratives. Not surprisingly, depressed women constantly use moral language, words such as "selfish," "bad," "worthless," as they assess themselves and their role in causing problems in their relationships. Gilligan's central metaphor was that of "voice," with self-silencing being the closest associated construct. Jack's Silence Subscale falls just over the F G border into HI (253°), but it is perhaps curious that no measure has ever been developed to expressly tap a "selfishness" construct as a unique form of self-punitive introject. The possible projection of a selfishness construct, spoken of by Jack and Gilligan, into F G warrants the development of such a construct. Regarding the possibility of further Octant F G and NO conventional care scales, further empirical work will be needed to confirm their existence and circular placement. Beyond the Self-directed Dimension From the interpersonal tradition (Sullivan, 1953), "interpersonal behavior is the ultimate defining medium in which 'personality' is manifested" (Carson, 1996, p. 242). The theorizations of Sullivan and other neo-Freudians (e.g., Adler, Fromm, Horney) have given greater weight to the complex and essentially social nature of human behavior than have earlier, more individualistic accounts. Leary's (1957) tacit principle of complementarity was made explicit with Carson's notions of reciprocity and correspondence (see above "Octant 115 NO (Gregarious-Extraverted)". In an article entitled '"Mental Illness' or Interpersonal Behavior," Adams (1964) expressed critical concern that the medical notion of "mental illness" in its literal organic sense was being used as a "verbal analogy .... applied to arbitrarily designated types of maladaptive interpersonal behavior" (pp. 191, 195). Adams contrasts the medicalization of psychological health care with "moral therapy" and the interpersonal approach he advocated: "impersonal scientism [is] quite different in its. basic outlook from the humanitarianism of the moral-therapy era [prior to the] .... 1870s and 1880s" (p. 193). On the eve of releasing the DSM-III McLemore and Benjamin (1979) would look back and signal the Adams article, similarly contending for an "interpersonal diagnosis" as an alternative to categorization. Since this time Interpersonal Therapy and Cognitive Behavioral Therapy have become the two most widely validated modes of clinical therapy (Anchin & Kiesler, 1982). Predominant among the several adaptations of Leary's (1957) interpersonal circle, Benjamin (1974, 1996) developed a microanalytic circumplex partitioned into 32s, the Structural Analysis of Social Behavior (SASB), which examines patterns of discourse between persons. Units of interpersonal exchange are scored on three dimensions: what have here been called self-directed, other-directed, and socially prescribed (see Figure 4 above) or in the SASB system "introject: focus inward," "transitive: focus outward," and "introject: action inward," respectively. These three dimensions reflect three parallel processes. For example, if your primary caregiver frequently attacked you, the impact can show up in one or more of three ways: (a) you become an attacking person (identification), (b) you become a person who fears being attacked, who easily feels attacked, or who chooses to be with attacking people (recapitulation); and (c) you attack yourself (introjection). (Benjamin, 1996, p. 251) With respect to the present study, the assumption of Gilligan (1982) and Jack (1991) is that (engendered) expectations to be nice, quiet, and caring will be internalized, and that the best way of assessing these is via the self-directed dimension (i.e., "In my friendships my 116 responsibility is to make the other person happy"). It may, however, also be true that rather than identifying with and internalizing role expectations, individuals may react against or respond in some other way to such expectations. Extending a broader concept of caring into the other dimensions may be of equal interest. First, a further note regarding the use of the same circular criteria used in developing and assessing self-directed scales is in order. The expectation that an analysis of item circular dispersion would prove of assistance was supported (Gurtman, 1993) in the case of self-directed items but was not expected to be evidenced across other dimensions. This was generally confirmed. Items from the other-directed dimension of the Social Perfectionism Scale (Wiebe & McCabe, 1998) and the Socially Prescribed Conventional Care Scale developed in the current study were broadly distributed. Items from the other-directed dimension of the Conventional Care Scale, however, were narrowly projected onto Octant DE. These findings should offer a caveat to those studies which have projected other interpersonal dimensions onto what is essentially a self-directed circumplex (e.g., projecting total self-oriented, other-oriented, and socially prescribed dimensions of perfectionism onto the IIP-C; Hi l l , Zrull, & Turlington, 1997). This is not to say, however, that the other two dimensions can not be projected onto interpersonal "trait" space, but only that the interpretation that ensues must be derived from a consideration of the overall distribution of items. Benjamin (1974) uses three distinct circumplexes for each of the three dimensions. Descriptive statistics and factor loadings for the resultant scales were reported above amongst the other Conventional Care Scales (see Tables 5 and 6 above). Circular projections of the three non-self-directed scales are given below (see Figure 20) - with the footnote that these projections represent a central tendency among a generally broader distribution of items relative to those examined thus far. The combined number of scales examined in the current study, then, involves nine self-directed scales discussed at length above, two other-directed scales, and one socially prescribed scale. 117 F i g u r e 20 Super Circumplex Scale Projections for the Other-directed Social Perfectionism, Conventional Care, and Socially Prescribed Conventional Care Scales (DOM + ) (PA) Assured-Dominant (DE) (HI) Unassured-Submissive Note. Above projections are for total scale scores. 1 = Other-directed Social Perfectionism (Wiebe & McCabe, 1998); 2= Other-directed Conventional Care; 3 = Socially Prescribed Conventional Care. To examine the relations and/or redundancy among the non-self-directed scales, a conjoint principal components analysis was conducted on the two other-directed and one socially prescribed scales. While an initial analysis revealed that four factors had eigenvalues greater than 1, the scree plot suggested the extraction of three factors. The analysis was repeated specifying a three-factor solution which accounted for 43.9% of the variance (see Table 9). Other-directed and Socially Prescribed Conventional Care loaded onto two distinct 118 Table 9 Factor Loadings for Non-self-directed Care Scales Factor Items grouped by scale 1 2 3 Other-directed Conventional Care 1 .48 .13 .32 2 • .58 .38 .23 3 .68 .03 .20 4 .63 .06 .27 5 .35 .30 .36 6 .66 .08 -.11 7 .62 .09 -.03 8 .60 .32 .12 9 .44 .04 .23 Other-directed Social Perfectionism 1 .22 .31 .19 3 -.03 .43 .28 5 .13 .19 .61 7 .27 .35 .15 9 .26 .15 .48 11 .14 .17 .32 13 .07 .02 .81 15 .04 .48 .24 17 .11 . .14 .60 19 .02 .13 .74 Socially Prescribed Conventional Care 1 .09 .54 .13 2 .15 .61 -.01 3 .23 .55 .12 4 .11 .54 .16 5 .15 .57 .07 6 .40 .49 -.03 7 -.08 .61 .06 Note. Loadings > .40 are indicated in bold. The wording of scale items for the Other-directed and Socially Prescribed Conventional Care Scales can be found in Table 5 and for the Social Perfectionism Scale in Appendix B. 119 factors. Social Perfectionism loaded primarily onto Factor 3 with two items loading greater than .40 on Factor 2, along with Socially Prescribed Conventional Care. The thematic content of Social Perfectionism items loading onto the third factor indicated a general concern that others live up to the respondents' expectations: "It is important that my friends live up to my expectations of them " (item 19), "To be worthy of my friendship, others should live up to my expectations" (item 13). Higher Factor 3 loading Social Perfectionism items, on the other hand, express a theme of critical disappointment: "My friends do things that disappoint me" (item 3) and "I am usually satisfied with the way my friends are towards me" (reverse item 7). It would appear that the Other-directed Social Perfectionism Scale is tapping two sets of social expectations, a critical disappointment in others factor (Factor 2) and a general concern factor (Factor 3). The scale's title "Social Perfectionism" suggests that high expectations are specifically being directed towards others on the dimension of nurturance, rather than agency. But it is unclear how these items are distinguishable from other already existing and well-validated scales which reflect this more general form of high expectation. Hewitt's Other-oriented Perfectionism Subscale (Hewitt & Flett, 1991) also contains a few general items such as "I have high expectations for the people who are important to me" (item 16). For the most part, however, Hewitt's measure and ones like it tap a more specifically agentic achievement form of expectation (e.g., "If I ask someone to do something, I expect it to be done flawlessly"). In subsequent analyses reported under the section "Gender Differences" the two subscales with the Social Perfectionism items do possess unique correlates. Examining the relation of all three scales within the Interpersonal Circumplex indicates some shared personality characteristics; all three scales' central tendency is to project into Octant D E (Cold-hearted). To this point, then, scales reflective of self-directed conventional care, including demand for approval, have been examined in relation to each other within a higher-order circumplex space. Additional dimensions of conventional care have also been examined in 120 their substantive relation to each other as well as in their relation to interpersonal space, with the above mentioned caveat regarding their broader item distribution. An examination of scales in circular space, particularly those within the same octant, can each inform an interpretation of the other. The relation of Silence, Anxious Concern, Demand for Approval, and self-criticism (from External) has already been briefly mentioned above. Doing so involves an investigation into what is referred to as convergent and discriminant validity, establishing each scale within a nomological network of other related and non-related scales. As significant is the use of the circumplex itself in establishing a higher-order structure in which to compare results from previous studies. It is to this second point that the discussion now turns. The Super Ci rcumplex as a Nomological Net: Octant D E (Cold-hearted) One of the advantages of a structural analysis is that it provides a comprehensive framework in which the results of the current study can be informed by previous studies -with the already mentioned footnote that angular projections will vary by "about 15° on average" (Gurtman, 1993, p. 259) across the IASR and the IIP-C. (For a visual representation of the relations between the two circumplex across their respective octants see Figure 6 and related discussion above.) It bears further reminding that interpersonal behaviors are viewed as occurring on a (circular) continuum and that their partitioning into octants (or 16s, etc.) is but a discursive heuristic. Research into using the IASR and IIP-C indicates meaningful differences at the octants level (Wiggins, 1995). Division of the circle into octants is only meant as an investigative tool whereby each octant can be viewed as an interpersonal prototype. Investigation of the various conventional care and dependency scales used in the current study can be given greater explication by examining their relation with other measures projected into the same octant which can be said to further define any particular octant. In any nomological network, however, the explanatory burden falls on the known constructs to explicate the less known. Newly devised scales are examined in their relation 121 with both conceptually related (convergent validity) and conceptually unrelated (discriminant validity) measures which have been already well-validated. Secondly, in addition to the horizontal nomological relation represented by the network metaphor, there is the vertical, or more specifically, hierarchical relation. According to Circumplex Proposition 2 above, constructs lower in the hierarchy will account for unique variance while nevertheless sharing common variance with their respective superordinate (or underlying) octant. Any projected scale, then, can be viewed (a) in relation to other scales within circular space, including the circumplex scales employed as a nomological net, but can also be viewed as (b) nested beneath their respective octants. The focus of the current study has been upon the "two faces" of self-directed conventional care, Octants HI and JK. Some preliminary forays were also made into Octants F G as well as NO. With the additional examination of non-self-directed dimensions of conventional care, Octant D E was included. The exploration of the lower half of the circumplex, then, will proceed counter-clockwise (DE to NO), excluding Octant L M for the reason that none of the present scales or preliminary scales fell within this region. Each octant will be discussed in relation to the results of previous studies employing the interpersonal circle, thus elucidating the interpersonal content of the respective octants. We start with the three non-self-directed measures projecting into Octant D E (Cold-hearted). Recall, however, that items from the socially-prescribed conventional care were more broadly spanning. Gilligan's (1982) argument is that young girls approaching puberty "dissociate" from their knowing and feeling, their "voice," by internalizing a "voice over." Young girls are met with conventional norm expectations to conform to the desirable stereotype of being nice, polite, pleasing to others, unassertive, and quiet - in short being a "good" girl. First, while the etiological role of family and culture, including peer culture, is central, the hypothesis has never been directed tested. This study is the first to propose the development of non-self-directed dimensions whereby this might be tested. Second, as interpersonal approaches 122 rightly note, socially prescribed expectations do not necessarily result in the identification with, and internalization of, such norms. There is ample evidence to suggest that the correlation need not necessarily be a high one (Hewitt & Flett, 1991). One response to divergent socially prescribed expectations is anger. Similarly, the one holding high expectations of others that they be a certain way, need not necessarily imply that they would meet their own expectations. Three non-self-directed measures fell within Octant D E (Cold-hearted) of "super" circumplex space: Other-directed Social Perfectionism and Conventional Care, and Socially Prescribed Conventional Care. The first two reflect the respondents' perception that significant others are falling short of their expectations of care, while the latter expresses the perception of falling short of others' expectations to be more caring. More specifically, six of the items within the Other-directed Social Perfectionism Scale fell within Octant D E and two fell in each of the adjacent Octants B C (Arrogant-Calculating) and F G (Aloof-Introverted). A l l nine items of the Other-directed Conventional Care Scale fell within DE, whereas only three of the seven items in the Socially Prescribed Conventional Care Scale projected D E with one item each in HI, PA, NO, and JK. Scoring high on any of these three measures is expressive of persons who tend not to be warm, cooperative, or caring; as reflective of their Octant D E projection. Rather, they emphasize freedom and autonomy. Studies examining the relation of this octant with a number of hostility scales such as the MMPI - Manifest Hostility Scale generally show only a small positive correlation - supporting Leary's (1957) observation that the hostility associated with this octant involves more covert attitudes of punishment and guilt provocation than overt acts of destruction. Other constructs that project into and further define Octant D E as measured by the IASR scales, are: Defendance, Authority Conflict, Machiavellianism, Cynicism, Suspiciousness, Paranoid Personality, Passive-aggression (Wiggins & Broughton, 1991), and Adult Dismissing Attachment (Bartholomew & Horowitz, 1991). The pattern of social exchange within this octant involves granting status but not love to self and the refusal 123 of both love and status to others. Hil l et al. (1997) projected Hewitt and Flett's (1991) three dimensions of perfectionism onto the Interpersonal Adjectives circumplex with the similar finding that the socially prescribed dimension projected onto Octant D E - however, they found that other-oriented perfectionism projected onto the B C (Arrogant-Calculating) octant. What is not clear is the question of where to place the weight of causal influence. Do individuals covertly and angrily withdraw in the face of unrealistic expectations that they be more achieving or caring? Or for some other reason were these individuals first underachieving and uncaring which elicited from others a strong press towards self-correction in the direction of greater achievement and caring? Given that the socially prescribed dimension reflects the perception of others' expectations, it is also possible that these perceptions reflect their own bias and that others in fact do not hold out such unrealistic expectations. The association between the Other-directed Conventional Care Scale and actual involvement and satisfaction in a relationship, and interpersonal competencies will later be examined in an effort to address this question. Anticipating the later discussion regarding gender differences, of the twelve care and dependency scales, only the Other-directed Conventional Care Scale evidenced any gender differences, with men reporting higher levels, F(l,300) = 5.28, p < .05. The only gender difference Hewitt and Flett (1991) report is on their other-directed dimension of perfectionism, with higher elevations for men. Men seem generally to place higher expectations of achievement onto others (i.e., other-directed perfectionism) and attribute their own lack of success to others' selfishness (other-directed conventional care). The Other-directed Conventional Care Scale seems specifically to reflect a sort of co-opting of the conventional moral language of selfishness and needs, which are employed to blame others (e.g. "If my friends weren't so selfish, I would be doing a lot better"). High scorers are quite demanding (e.g., "Other people should be willing to interrupt their plans to make room for my needs"). A number of items make reference to the respondents' "needs" 124 (e.g., "My friends should just realize how much I am in need"), although it is not clear whether the blame and anger arise from the respondents' actual awareness of dependency needs or whether the reference is used as a secondary justification. Put differently, is the Other-directed Conventional Care Scale a direct measure of dependency or of moral blaming? Borrowing from the attachment literature some speculations can be drawn. The dismissing attachment category projects onto this same cold-hearted octant of the IIP-C. Within Bartholomew's four-category model of attachment (Bartholomew & Horowitz, 1991), individuals who are categorized as Dismissing are understood as occupying one quadrant of a two-by-two table: model of self (as worthy of love, or not) by model of other (as trustworthy to provide love, or not). Dismissing individuals are those who report themselves positively and others negatively in this regard. With regard to the self, however, some theorists believe that conscious report is a guise for elevations of actual unexpressed (nonconscious) dependency needs. The counterpart in the infancy literature, for example, finds that dismissing infants seem behaviorally unperturbed by the mother's absence, redirecting themselves towards objects. Yet, when physiological measures are taken, these dismissing infants tend to become more aroused than infants who do show external signs of distress. These findings have been replicated among adults who report confidence and a lack of interpersonal anxiety but whose bodily physiology seems to indicate considerable distress. With reference to the current study, however, these must remain exploratory references. The Super Circumplex as a Nomological Net: Octant F G (Aloof-Introverted) While only a few provisional items projected into Octant F G (Aloof-Introverted), the possibility that a "selfishness" construct of clinical importance resides here warrants some further elaboration of the octant. Two items in the current study locatable in F G were: "To look after my own needs is completely selfish" and "Taking better care of myself would be selfish." Consistent with previous studies (Wiggins & Broughton, 1991), both Rosenberg's Self-esteem Scale (Rosenberg, 1965) and Beck's Depression Inventory (Beck, 1967) projected into Octant FG, reflective of the high zero-order correlations with the octant's scales. 125 Moderate correlations were found predicting depression and self-esteem; IASR Octant F G scale (rs = .35 and -.46, ps < .001, respectively) and the IIP-C Octant F G (rs = .23 and -.31, ps < .001, respectively). Individuals whose peak elevation is in Octant F G are generally characterized by a tendency to avoid social interactions and to reject the warm gestures of others. They describe themselves as introverted, aloof, distant and unsociable, and use a variety of strategies, such as turning down invitations and avoiding interpersonal contact, as means of limiting social life. The Super Circumplex as a Nomological Net: Octant HI (Unassured-Submissive) Three scales projected into Octant HI (Unassured-Submissive): Silence, Self-directed Social Perfectionism, and Anxious Concern. A l l three scales share an anxious concern in contexts of social exchange. Consistent with the points made above, individuals who score high in Octant HI have a tendency to be timid, fearful, and submissive in interpersonal transactions. Individuals who report high scores describe themselves as meek, timid, shy, and self-doubting. Such individuals avoid being the center of attention and social situations involving challenge and authoritarian power over others. Measures projecting onto Octant HI include Trait Anxiety, Dependent Personality, Fear of Negative Evaluation, along with several measures developed from Murray's (1938) taxonomy, Infavoidance (i.e., avoiding actions or situations where failure might result) and Succourance (i.e., seeking love and assistance) (Wiggins & Broughton, 1991). The octant generally is expressive of difficulty in making one's own needs known and a difficulty with self-assertion. In terms of social exchange, Octant HI reflects a pattern in which love and status are denied the self, and status but not love is granted to the other. This is consistent with the projection of Fearful Attachment into this octant (Bartholomew & Horowitz, 1991). Fearfully attached individuals report negative models of both self and others with regard to love. The Silencing the Self Subscales (i.e., Care, Silence, External, Divided) have been 126 interpreted by their users (e.g., Thompson, 1995) as of a single piece, in contrast to the current finding that suggests two distinct underlying interpersonal styles defined by Octants HI and JK. Primarily what distinguishes the two octants in exchange theory (Carson, 1969) is the granting of love to the other, found in Octant JK and absent from Octant HI. Both octants grant status to the other and neither status nor love to the self. A second point involves how results from the Silencing the Self Scale have been interpreted in light of the notion of anxious (i.e., preoccupied) attachment - involving a desperate emotional clinging to an attachment figure, fear of being alone, and so forth. The External and Silence subscales, however, are associated more closely with fearful attachment than preoccupied attachment. A comparison of the two attachment styles can be grasped from the two paragraphs below. In Bartholomew and Horowitz's (1991) self-report Relationship Questionnaire participants are asked to rate themselves on the degree to which they resemble each of four attachment styles. The Preoccupied and Fearful attachment styles are described below. Preoccupied. I want to be completely emotionally intimate with others, but I often find that others are reluctant to get as close as I would like. I am uncomfortable being without close relationships, but I sometimes worry that others don't value me as much as I value them. Fearful. I am uncomfortable getting close to others. I want emotionally close relationships, but I find it difficult to trust others completely, or to depend on them. I worry that I will be hurt if I allow myself to become too close to others, (p. 244) While the domain "when I am with my friends" is stated within the actual scale items, the proximal motive implied by Jack (1991), that of "maintaining relationships," is not. From attachment and interpersonal theory, however, many forms of complementarity or maintenance of relationships exist. For a spouse caught in a violent relationship, "maintaining relationships" might mean only ensuring the continuation of one's economic means of support, which may or may not be intertwined with other attachment concerns. Consider, for example, an interview I had with an individual who as a child was 127 raised in an abusive family situation. This woman reported engaging in many self-silencing and -sacrificial caring behaviors, but not out of any gender-specific familial expectations. Rather, these were engaged in for the purpose of retaining sufficient stability in the family until she could leave. An individual can self-silence because she or he is afraid of negative evaluation and seeks approval from others. In moving too rapidly from the first point, women's interdependent self-construal, to the second, an ethic of care, Gilligan's theory risks over-generalizing and constraining alternate hypotheses. One reinterpretation of Gilligan's "care" ethic is that it represents a "slave morality." Puka (1993) interprets Gilligan's ethic of care as a blue-collar orientation toward domination or a "third world" orientation to the "economic imperialism" of industrial countries. Reinterpreted, Gilligan's Level 2 survival strategy states, to overcome ongoing powerlessness, play the roles those in power set for you. Serve and sacrifice to gain their approval and support, thereby participating in their power and avoiding harm. Be circumspect in pursuing your true interests, or even in recognizing them. (Puka, 1993, p. 223) That the IBT - Demand for Approval Scale, used in the present study, along with the Fear of Negative Evaluation Scale (Watson & Friend, 1969) employed in previous studies (Wiggins & Broughton, 1991), project into the HI octant should provide a clue that the Silence Scale is tapping a more avoidant motivation reflective of a negative internal model of others as critical. Puka's interpretation employs the "coping" metaphor with its transient contextualism. What gets neglected is the role that an underlying conventional stage of cognitive development might play. The Super Circumplex as a Nomological Net: Octant JK (Unassuming-Ingenuous) It has here been argued that an ethic of care can be understood relative to various facets, particularly Octants HI and JK, which are being proposed as the "two faces" of conventional care. Five care scales fell within Octant HI, including the (dependency) IBT -Demand for Approval Scale. Four scales projected into the superordinate circumplex space defined by Octant JK (Unassuming-Ingenuous): Care, Unmitigated Communion, and two 128 care scales developed in the current study, Self-Sacrificial Care and Perfectionistic Care. These scales generally reflect the primary emphases in the care literature - care is understood as self-sacrificial and the needs of others are deemed more important than one's own. The addition of the Perfectionistic Care Scale is a more direct expression of self-standards regarding care and has thus far not been reflected in any previous measures. The underlying octant itself reflects a tendency for respondents to be deferent, obliging, and modest. Descriptors on the Octant JK Interpersonal Adjectives Scales (IASR) describe themselves as mild, gentle, and conventional, not argumentative or egotistical. Measures which project onto this octant include: Deference, Abasement, and Self-control (Wiggins & Broughton, 1991). For example, a typical item from Jackson's Abasement Scale (1987) reads, "Several people have taken advantage of me but I always take it like a good sport." It is really only with this "second face," Octant JK, that the notion of "care" as a descriptor seems more apt. Octant JK is a "blend" of communion and submissiveness. Where "care goes awry," here, is in leaving the self out of the picture; in terms of social exchange (Carson, 1969), while granting care and status to others, both are denied the self. Participants whose profile peaks in Octant FU are prototypically regarded as not being able to grant love to others. What links the two faces under the heading of conventional care is the way in which conventional moral language is employed. Persons within Octant FU may want close relationships and construe their anxious worry about interpersonal interactions in terms of being "uncaring" or "insensitive," but the relationships they maintain are likely to be more distant and cautionary. The Super Circumplex as a Nomological Net: Octant NO (Gregarious-Extraverted) While the focus of the current study was upon Octants FU and JK, some preliminary investigations of items from within the provisional pool of conventional care suggested the relevance of Octants F G and NO to an ethic of care. Two preliminary scales emerged, entitled Selfishness and Intrusive Care. While considerable research has been done on guilt and shame (e.g., Tagney, Wagner, & Gramzow, 1992), it is perhaps surprising that the moral 129 experience of "selfishness" has not received more empirical investigation. Examining the thematic contents of the IIP-C Octant NO indicates four distinct thematic facets: excessive self-disclosure, exhibitionism/wanting to be noticed too much, anxiety about being alone, and of current interest, two intrusive items: • I feel too responsible for solving other people's problems • Hard to stay out of other people's business It would appear that when a high degree of concern for others is combined with higher levels of agency than that present towards the submissive pole, a form of interpersonal intrusiveness emerges. While only preliminary, this would lend some support to Birtchnell's (1987) suggestion that Bowlby's notion of compulsive care-giving be linked with the upper right quadrant. In terms of the circumplex, this represents a blend of dominance and communion, an over-bearing sort of concern for others. In addition to the greater agentic involvement of care within Octant NO, compared to Octant JK, a second distinction that can be hypothesized is the former's more practical orientation. In addition to a variety of sociability and gregarious scales projecting into Octant NO, Wiggins and Broughton (1991) report Murray's (1938) Exocathection (i.e., "interested in practical activity and the affairs of everyday life.... [versus Endocathection] interested in things of the mind" p. 222) as falling within the octant as well. Dramatic/ histrionic elements are also reported here. In summary, projecting scales into circular space allows for their examination within a theoretic-empirical framework which can be employed as a nomological net - in contrast to the all too frequently employed dust-bowl empiricism reflective of atheoretic applications of convergent and divergent validity approaches. The present investigative "mapping" of the care domain was begun at a micro level, investigating the structural projections and thematic content of individual scale items. How items load and relate to one another was examined by using a conjoint principal components analysis. Further, the resultant six extracted factors, as expressed by their representative scales, were located in two octants, HI and JK. While the 130 focus has been upon the two faces of Octants FU and JK, preliminary scales were also suggest for Octants F G and NO. Moving outward in an ever macroanalytic direction, the scales were examined in their nomological relation to interpersonal space generally. The advantage of a superordinate factor space is to provide a common language, so to speak, whereby scales from divergent research traditions can be discussed and compared. For example, Silencing the Self (STSS; Jack, 1991) has been largely examined as a unitary construct producing a global score. But as the current research demonstrates, apart from the Silence Subscale, neither the global construct nor the scales themselves sample from a singular space of semantic meaning. Gilligan has too narrowly constrained the range of interpretive meaning that might be given to the results of studies employing the scale. One such misinterpretation involves the linking of self-silencing (as a global construct) with "anxious attachment." Proposing this link risks making at least two errors. One is to confuse the interpersonal domain of friendships, which Gilligan speaks most about, with attachment domains involving intimate relationships. Thompson's (1995) study involved an investigation of the STSS scales within intimate relationships. But as Harter et al. (1998) have demonstrated, the construct of "voice" functions differently within close intimate relationships than it does in relationships with peer, parents, and teachers. The second error in linking the STSS with adult attachment brings us back to the multifaceted nature of conventional care. Perhaps surprisingly, neither existing scales nor the a priori scales developed in the current study fell within Octant L M - precisely where anxious (preoccupied) attachment falls (Bartholomew & Horowitz, 1991). Linking conventional care, with anxious attachment demonstrates the operating role of gender stereotypes which are here serving only to constrain alternate hypotheses. While it is true that studies of attachment across several self-report questionnaires have evidenced a gender difference in favor of women reporting more anxious attachment and anxious concern with intimate relationships generally (Shaver & Clark, 1994), these findings have not been consistent. No gender differences are found when using the attachment interview (Bartholomew, 1994) These 131 attachment constructs are actually orthogonal to the two STSS scales located in Octant HI. Too hastily, a variety of constructs that span a vast array of interpersonal meaning are simply slung together. Jack (1991) termed the socially approved collection of attachment behaviors for women "compliant connectedness" (p. 40).... which resembles anxious attachment, is characterized by compulsive caretaking, pleasing the other, and inhibition of self-expression. (Thompson, 1995, p. 338) Similarly, Gilligan's Level 2 care, like Kohlberg's conventional Stage 3, makes too much of the assumption of a desire to "maintain relationships," where on the obligation (versus entitlement) side this means something like anxious attachment or simply a desire for the other's good. In the next series of analyses, conventional care scales are examined in their relation to reported interpersonal capabilities. Interpersonal Competencies Correlational analyses were undertaken to examine the question of the various care scales' relation to interpersonal competencies. That Silence, for example, falls within Octant HI already suggests that high scorers are likely to have poor interpersonal skills, more so than high scoring individuals on scales within Octant JK. This next section might then be thought of as further confirming these findings from the circumplex using scales developed within the social skills field. Buhrmester and colleagues (1988) identified five domains of interpersonal competencies (i.e., social skills) seen as central in effective social interactions. Three of the scales were employed in the current study: Negative Assertion (e.g., "Telling a companion when he or she has done something to hurt your feelings"), Emotional Support (e.g., "Being a good and sensitive listener for a companion who is upset"), and Self-disclosure (e.g., "Telling a close companion things about yourself that you're ashamed of"). (Recall that Negative Assertion was already shown above to be distinct from the silence construct.) The projection of these scales into interpersonal space can readily be predicted from the Interpersonal Circumplex. Negative Assertion is a close opposite of Silence and the HI octant generally 132 and thus should project into PA (Assured-Dominant). The items that comprise the Negative Assertion Scale do fall narrowly within PA (CM = 87°, SCD = 6.3°), as reported above. Emotional Support could also be anticipated to project within L M (Warm-Agreeable), which all of its items do (CM = 12°, SCD =11.1°). The Self-disclosure Scale could be anticipated to either reflect components of L M or the more self-disclosing aspects of NO. An item from the IIP-C NO scale reads: "I open up to other people too much." A l l items within the scale were found to project narrowly within NO (Gregarious-Extraverted: CM = 39°, SCD = 10.8°). Zero-order correlations of care and dependency scales will be set out by octant (Table 10). As could be predicted from Circumplex Proposition 1, those scales opposite in octant within the Interpersonal Circumplex were the most highly (negatively) correlated, whereas those scales projecting orthogonally possessed low non-significant correlations. For example, Silence was negatively correlated with Octant PA Negative Assertion Scale (r = -.46, p < .001), moderately correlated with Octant NO Emotional Support (r = -.35, p < .001), and correlated weakly with Octant L M Emotional Support (r = -.19, p < .001). Scales within Octant HI were largely non-significant with the orthogonal Octant L M . Similarly, moving counter-clockwise from Octant HI to JK results in a change in correlations from essentially zero to moderate with Octant scales L M (Emotional Support), and a decrease in the negative relation with scales in Octant PA (Negative Assertion). Octant NO is now orthogonal to JK and results in non-significant correlations between scales within the two octants (Self-disclosure and JK scales). The relationship is expected to be less evident for the non-self-directed scales. The confirmation of this in Table 10 supports the caveat regarding the use of the Interpersonal Circumplex in the examination of non-self-directed scales. Octant HI summary. A quick summary of the above relations can easily be grasped from their graphic representation, along with relations to the Five-Factor Model yet to be discussed, in Figures 21 to 23 below. Generally and as expected, all Octant HI scales predict 133 Table 10 Correlations between Care/Dependency Scales and Interpersonal Competencies Interpersonal Competencies Negative Emotional Self-Assertion Support Disclosure Scale (Octant PA) (Octant L M ) (Octant NO) Octant HI (Unassured-Submissive) Silence -.46 -.19 -.35 External -.37 -.10 -.24 Social Perfectionism - Self-directed -.33 -.09 -.22 Anxious Concern -.43 -.09 -.19 Demand for Approval -.42 -.02 -.18 Octant JK (Unassuming-Ingenuous) C a r e -.29 -.07 - 14 Unmitigated Communion __2l 24 01 Perfectionistic Care _ 19 26 12 Self-Sacrificial Care . 19 0 9 Q 3 Octant D E (Cold-hearted) Social Perfectionism - Other-directed -.07 -.08 -.17 Conventional Care - Other-directed -.16 -.21 -.13 Conventional Care - Socially Prescribed -.16 103 -.13 Note. Correlation coefficients greater than .11 are significant at/? < .05, those greater than .15 are significant at p < .01, and those greater than .18, p < .001. problems in the area of Negative Assertion and to some extent Emotional Support for the Silence and External Scales. Self-disclosure is also low for all scales, particularly the Silence scale. Of all the scales, Silence is the greatest predictor of problems in the domain of ^'interpersonal capabilities. Octant JK summary. In comparison with Octant HI scales, the Octant JK scales predict less difficulty with negative assertion (low to moderate correlations), but in considerable contrast do predict, in the case of Perfectionistic Care and Unmitigated Communion, emotional support. The case also improves for self-disclosure, with non-significant to low relations for care scales. High scorers among the JK scales are warmer but 134 continue to have difficulty in letting their needs and wants be known. Amongst the JK scales, Perfectionistic Care predicts the highest levels of emotional support and self-disclosure. Self-disclosure involves an ingratiating bid for reciprocal intimate involvement, fostering the impression that a person is honest and likes the other (Tedeshi & Melburg, 1984). Octant DE summary. The non-self-directed scales (Figure 23 below) are generally not predictive of interpersonal capabilities. The results of this study support the hypothesis of at least two "faces" of conventional care. Consistently, scales falling within Octant FU have been most closely associated with problems, including interpersonal problems. Together, these analyses provide more detail to the comment that conventional care is about some diffuse "maintaining of relations." Additionally, a second component of interpersonal functioning was assessed, moving from the broader domain of friends, assessed by the interpersonal competencies investigated above, to the domain of more intimate relations. Participants were asked on the demographics form about their intimate relationships: whether the participant was currently in a relationship, and if so for how long. Participants were also asked to rate their level of satisfaction and their perceptions of their partner's level of satisfaction with the relationship. Over one third of the sample (N= 111) reported currently being in an "intimate relationship." The mean length of the relationship was approximately one-and-a-half years (M = 18.7 mths, SD = 15.7 mths, range = 1 - 66). A series of one-way A N O V A s was conducted examining differences in the three interpersonal competencies scales. Those in an intimate relationship reported higher levels than those not in a relationship across all three interpersonal competencies: ICQ - Negative Assertion, ICQ - Emotional Support, and ICQ - Self-disclosure, F(l,297) = 15.05, 4.20, and 10.11; ps < .001, .05, and .005, respectively. While the possession of a certain level of interpersonal competency is predictive of being in a relationship, it is likely also true that the context of relationships is a training ground for the further refinement of skills already present. Partialling out age, given its correlation with ( 135 length of relationship (r = .40), only the Negative Assertion Scale was found to be predictive of the actual length of the relationship (r = .24, p < .05). Participants who reported being in an intimate relationship were also asked to report their satisfaction with the relationship as well as the partner's (perceived) satisfaction. A measure of mutual satisfaction was created from the product of self- and partner-satisfaction ratings. Negative Assertion and Self-disclosure predicted level of self-satisfaction with the relationship (rs = .19 and .30; ps < .05 and .01, respectively), but only Self-disclosure predicted mutual satisfaction (r = .27, p < .01). Ratings of self-satisfaction and (perceived) partner-satisfaction were strongly correlated (r = .71, p < .001). Similar to the findings of other studies, the Interpersonal Competencies Questionnaire scales do predict actual interpersonal involvement in and satisfaction within intimate relationships. While interpersonal competencies predicted relational satisfaction generally, conventional care scales generally did not. Of the nine care/dependency self-directed scales, only Silence was (negatively) predictive of self- and (perceived) partner-satisfaction with the relationship (rs = -.19 and -.20, ps < .05). Both Other-directed Conventional Care and Social Perfectionism (negatively) predicted self-satisfaction (rs = -.21 and -A9,ps < .05). So while it appears that there exists a strong relation between care/dependency scales and interpersonal competence, and between interpersonal competence and actual satisfaction within relationships, it appears that a the path from the former to the later may be a mediated one. Several alternate interpretations suggest themselves. A distinction, following comments made above, could be made between intimate versus social/peer domains. A person could assert himself/herself within intimate relationships but find these same interpersonal capacities difficult within friendship/peer relationships. Or, persons scoring high on Octant HI scales, for example, could also have very low expectations for what satisfaction in a relation might be. They report problems in self-confidence, self-assertion, demand others' approval, and are anxious about social interactions, yet report being satisfied. Further studies will need to ferret out these distinctions. 136 Beyond the Circumplex A more comprehensive picture of care and dependency can be gained when their analyses are extended beyond the space of the Interpersonal Circumplex to include the remaining factors of the Five-Factor Model: Neuroticism, Conscientiousness, and Openness (recall that the circumplex plane corresponds to Agreeableness and Extraversion in the Five-Factor Model). It is important to go beyond the circumplex for the reason that even where scales may occupy similar locations in circular space they may possess divergent patterns of correlations with respect to other super factors, adding distinct nonpersonal nuances to their interpretation (Circumplex Proposition 2). Of .particular concern are those situations in which intercorrelations among scales are inflated owing to shared distress (e.g., Neuroticism) variance - thus masking possible distinct underlying interpersonal behaviors. This has already been noted above in discussion of the possibility that individual items may survive screening for inclusion in parent scales owing to a shared distress factor. Looking across the self-directed scales, reported in Figures 21 and 22, care/dependency scales have their strongest correlations with interpersonal factors (DOM/NUR) and Neuroticism. The HI scales as a group were most strongly interpersonal, along with JK's Perfectionistic Care Scale. This was less so of the non-self-directed scales, particularly for the Other-directed Conventional Care Scale (see Figure 23). Of the three octants (DE, HI, JK), HI scales were most strongly associated with Neuroticism, although perhaps surprisingly, only a low-moderate association was found with Silence. D E and JK scales were moderately correlated with Neuroticism, although Perfectionistic Care was the least elevated among them. Across all scales Conscientiousness and Openness were non-significant, again with the exception of the Silence scale. Generally, then, self-directed scales were marked by moderate to high Interpersonal and Neuroticism factors. Perfectionistic Care possessed a less problematic profile involving lower Neuroticism, moderate Conscientiousness, and extending to interpersonal competencies, moderate to high Self-disclosure and Emotional Support. Some difficulties with Negative 137 Figure 21 Personality and Interpersonal Correlates of the Care and Dependency Scales for Octant HI ( Unassured-Submissive ) i8_ . • 6H -.6 DOM/NUR C O N NEUR OPEN NA ES SD SUPER FACTORS INTERPERSONAL COMPETENCIES -o— Sel f -directed Social Perfect ionism (SPS) - • - S i l e n c e (STSS) —A—Anxious C o n c e r n (CCS) —o— D e m a n d for Approva l (IBT) - • - E x t e r n a l (STSS) Note. DOM/NUR = Dominance/Nurturance; C O N = Conscientiousness; NEUR = Neuroticism; OPEN = Openness to Experience; N A = Negative Assertion; ES = Emotional Support; SD = Self-disclosure. Assertion, however, are associated with Perfectionistic Care. Silence was a second distinct scale, reporting lower associations with Neuroticism and moderate negative correlations with Openness. The factors of Conscientiousness and Openness are particularly important if individuals are to garner the motivational and cognitive resources needed to benefit from therapeutic or self-efficacy experiences. A question arises for the Silence Subscale involving the lower than expected correlations with Neuroticism. Should this be regarded as the true state of affairs, or does the simultaneous low Openness contribute to a lack of awareness of one's own 138 Figure 22 Personality and Interpersonal Correlates of the Care and Dependency Scales for Octant JK ( Unassuming-Ingenuous ) .8 -, 1 _ - 6 J . DOM/NUR C O N NEUR OPEN NA ES SD SUPER FACTORS INTERPERSONAL COMPETENCIES - o - C a r e (STSS) — • — U n m i t i g a t e d C o m m u n i o n (UCS) — A — Self-sacrificial C a r e (CCS) -o— Perfectionistic C a r e (CCS) Note. DOM/NUR = Dominance/Nurturance; C O N = Conscientiousness; NEUR = Neuroticism; OPEN = Openness to Experience; N A = Negative Assertion; ES = Emotional Support; SD = Self-disclosure. internal processes? Are high-scoring self-silencers repressors (Weinberger, 1990)? Development of care scales based on Gilligan's (1982) Level 2 ethic of care was intended to address issues that are more prevalent among women, such as depression (Jack, 1991). Similarly, their inclusion in the study reported in the following chapter was thought to shed some light on eating disordered behaviors in women. Twice as many women suffer from depression as do men, and eating disordered behavior remains a social phenomenon largely limited to middle- to upper-middle-class white women. The relation between the care scales 139 Figure 23 Personality and Interpersonal Correlates of the Care and Dependency Scales for Octant DE ( Cold-hearted) .8 -i 1 • 6 A -.4 --.6 • : DOM/NUR C O N NEUR OPEN NA ES SD SUPER FACTORS INTERPERSONAL COMPETENCIES -o— Other -d i rec ted Social Perfect ionism (SPS) - • - S o c i a l l y Prescribed C a r e (CCS) — A— Other -d i rec ted Conven t iona l C a r e (CCS) Note. DOM/NUR = Dominance/Nurturance; CON = Conscientiousness; NEUR = Neuroticism; OPEN = Openness to Experience; NA = Negative Assertion; ES = Emotional Support; SD = Self-disclosure. and measures of well-being (depression and low self-esteem) will be examined subsequent to the question of gender differences which is undertaken next. Gender Differences: Factor Structure Measures Gender differences across several of the various measures employed in the study were found and will be discussed following something of the order of their introduction in the "Method" section above. First, gender differences in the factor structure measures, the IASR-B5 and the IIP-C, will be discussed. In the following section gender differences 140 among the care/dependency measures and interpersonal competencies will be reviewed, and finally, two specific outcome measures will be examined, the True/False-self and the care narratives. We turn first to the factor structure measures. Gender differences across the eight octant scales for each of the IASR-B5 and IIP-C octant scales were asssessed using univariate tests of significance. The ipsatization of IIP-C scales results in a degrees-of-freedom problem when conducting between-subjects analyses such as with multivariate tests; this is not a problem in conducting principal components analyses, which do not involve the calculation of an inverse from the intercorrelation matrix. The additional three super factors of the IASR-B5 (Conscientiousness, Neuroticism, and Openness) were also examined. Univariate tests revealed significant gender differences across octant scales that represent generally the hostile-dominant versus the friendly-submissive quadrants, with women's profiles peaking in the latter and men's profiles in the former. More specifically, and consistent with previous studies, the IASR group profiles for females indicated significant peak elevations on Octants JK (Unassuming-Submissive) and L M (Warm-Agreeable), F(l,290) = 22.58 and 8.50, ps < .001 and .01, respectively. On the IIP-C, women reported more interpersonal problems in Octants HI and JK, F( 1,290) = 7.81 and 14.77, ps < .01 and .001, respectively. The men's group profile reflects the mirror image of the female profile, with significant peak elevations on IASR Octants B C (Arrogant-Calculating) and D E (Cold-hearted), F(l,290) = 29.96 and 17.43, respectively, ps < .001. The IASR group profiles are graphically illustrated below in T-score form (see Figures 24 and 25, for women and men, respectively). On the IIP-C, men reported more interpersonal problems in Octants B C and DE, F(l,290) = 29.96 and 17.43, ps < .001, respectively. There is some evidence to suggest that gender-role-related self-presentation may not reflect gender differences in empathic responsiveness or nurturant, altruistic, or agreeable behavior (Eisenberg & Lennon, 1983; Graziano & Eisenberg, 1997). Mitigating against a self-presentational interpretation are the findings of consistency with peer ratings (Wiggins, 141 Figure 24 IASR Profiles for Women (DOM + ) (PA) Assured-Dominant 60 (HI) Unassured-Submissive Note. Octant scores are indicated in T-score format based on combined male and female group norms for university samples (M = 50, SD = 10; Wiggins, 1995). 1995) and the similar findings for the IIP-C, suggesting that these gender differences are likely "true" profile differences. The interpersonal circumplex approach has already been shown to have given clarity to the constructs of "masculinity" and "femininity," reinterpreting them in terms of dominance and nurturance, thus avoiding the confounding of gender stereotypes. When Wiggins and Holzmuller (1978) removed three items from the Masculinity (M) Scale (athletic, individualistic, and masculine) and four items from the Femininity (F) Scale (childlike, shy, flatterable, gullible) of Bern's Sex Role Inventory (BSRI; Bern, 1974) they found a correlation of .83 between D O M and M , and a correlation of .89 between NUR and 142 Figure 25 IASR Profiles for Men (DOM + ) (PA) Assured-Dominant 60 (HI) Unassured-Submissive Note. Octant scores are indicated in T-score format based on combined male and female group norms for university samples (M = 50, SD = 10; Wiggins, 1995). F. The BSRI narrowly focuses on two octants only, rather than the full eight which the IASR measures. The advantage of investigating personality from the perspective of a circumplex is the recognition that interpersonal behaviors can conceivably vary the full 360°. Extending beyond interpersonal space, gender differences are also frequently found in neuroticism, with women reporting greater negative affect than men. These differences are found in spouse ratings as well as in self-reports, and may not be entirely attributable to sex bias in the willingness to report negative affect (Costa & McCrae, 1985). In the current study, no gender difference in Neuroticism was found, but women reported higher levels of Conscientiousness, F(l,290) = 7.04, p < .01. On the factor of Openness, no gender 143 difference was found. Across both the IIP-C and the IASR (less Neuroticism, Openness, and Conscientiousness), then, men and women do describe themselves interpersonally and disclose interpersonal problems consistent with gender stereotypes. Women describe themselves using interpersonal traits such as "tenderhearted," "accommodating," and "undemanding," and report greater problems in these same areas, such as "trying to please others too much" and finding it "hard to attend to my own welfare when somebody else is needy." Men, in contrast, describe themselves interpersonally using terms like "hardhearted," "ruthless," "calculating," and report related interpersonal problems such as finding it "hard to trust other people," "hard to really care about another person's problems," and "hard to feel close to other people." Gender Differences: Care/dependency and Outcome Measures A single M A N O V A was conducted across all the (conventional) care and dependency scales, as well as the outcome measures, interpersonal competencies, True/False-self questions, depression, and self-esteem. The initial M A N O V A was significant, F(19, 275) = 2.91, p < .001. Follow-up univariate tests revealed significant gender differences on four variables: Other-directed Conventional Care Scale (CCS), Emotional Support (ICQ-ES), and both of the True/False-self questions. Regarding the Other-directed Conventional Care Scale (CCS), men reported greater demandingness of others and attributed their lack of received help to friends' selfishness. This is also consistent with the above picture of men, both describing themselves in terms of, and reporting problems in, the hostile-dominant Quadrant I, F(l,290) = 6.44, p < .05 Somewhat surprisingly for a theory that takes its beginnings from the assumption of gender differences, differences were not found on any of the self-directed care/dependency measures. This might also be surprising given the strong differences for women both in peak Quadrant IV self-descriptions and interpersonal problems seen in the profiles above. These findings are consistent, however, with other studies using Jack's (1991) Silencing the Self Scale: either no gender differences are found, or men are reported as indicating higher levels 144 of global self-silencing, aggregating across all of Jack's subscales (Thompson, 1995). Of the outcome measures, only on the Emotional Support Scale were differences found, with women reporting a greater capacity to listen, understand, show concern and empathic support for friends (i.e., the stereotypic female virtues), F(l,290) = 14.56, p < .001. In the face of an absence of gender differences in Jack's (1991) Silencing the Self Scale, investigators suggest that what is of significance is the pattern of correlations with predictor variables such as satisfaction within relationships and depression. While rates of depression for women are twice those of men, there has been a lack of clarity regarding an explanation for this difference (Nolen-Hoeksema, 1990). It should be noted that no gender differences were found in levels of depression in the current sample. This could be attributed to the generally higher level of functioning of university undergraduates in comparison with community samples. However, in the current sample, there was a wide range of scores, including 7.3% above threshold for severe depression. Thompson (1995) found that despite men's greater global self-silencing (i.e., summed global score across all Jack's Silencing the Self Scales), it was only women's self-silencing that predicted relationship satisfaction, both for themselves and their perceptions for their partner. Thompson reports the same divergent pattern of correlations predicting levels of depression, with a significantly stronger relationship existing for women than for men. The first of these hypotheses, that self-silencing is more negatively predictive of relationship satisfaction for women than for men, was tested in the current study by asking participants who reported being in a relationship (N= 111) to indicate their level and their perception for their partner's level of satisfaction with the relationship (see above section "Interpersonal Competencies" for more detail). For each of the pairs of correlations, comparisons across gender (Fisher r to z transformations) were computed. Across the self-directed scales, Octant HI Silence and Anxious Concern Scales did differentially predict reported satisfaction with the relationship by gender (zs = 2.03 and 2.10, respectively, ps < .05), but in the opposite direction from Thompson's findings. Silencing 145 men reported a stronger negative correlation with relational satisfaction (r = -.32, p < .001) than did women (r = .08, n.s.). Creating a global score across the three Silencing the Self Scales used in the present study, following Thompson (1995), indicated similar results with the addition of a gender difference on mutual satisfaction, coded as the sum of self-satisfaction and perceived partner-satisfaction. Men (rs = -.33 and -.30, ps < .001) reported a significantly stronger negative correlation than did women (rs = -.08 and -.07, n.s.) for both self-satisfaction and mutual satisfaction in the relationship (zs = 2.21 and 1.96, ps < .05). No relation was found for perceived partner's satisfaction. It should, however, be noted that the two relationship items used in the present study do not possess the reliability of the Dyadic Adjustment Scale (Spanier, 1976), nor are dating undergraduate students comparable to the older and more relationally committed couples in Thompson's study. Regarding the hypothesis that self-silencing women report greater levels of depression than do men, contrary to Thompson's (1995) findings, no significant gender differences were found on any of the care scales, including the global self-silence scale. Thompson (1995) gives no account of what hypotheses her findings of a significantly stronger relation between global self-silencing and depression/relational satisfaction for women than men are supposed to support, nor are any post hoc explanations given about these differences. The discussion of what Gilligan's theory would predict, along with equally justifiable alternate hypotheses, will be taken up in the last chapter. What such a discussion might include is the notion of a false-self, particularly given the finding in the present study that men reported higher levels of false-self orientation than did women. Given the interpretive centrality of the false-self, it will be discussed within a separate section. The subsequent section, "Gender Differences: Motives for Helping" will likewise be discussed under the general topic of gender differences. Gender Differences: False-self Harter (1986) has been pursuing the question of why girls' levels of self-esteem declines after Grade 3, whereas self-esteem in boys remains relatively stable. Women, 146 particularly conventional women in the role of homemaker, have been found to report lower levels of self-esteem than men (Harter, 1992). Recall that Gilligan's theory holds that young girls repress and dissociate from their "true" selves to take up the voice of the detached other (i.e., patriarchy). Women's true self includes the value of relationships, caring, and cooperation rather than detached competition. Concurrent with the decline of self-esteem amongst girls in early adolescence is the developmental emergence of a false-self. Individuals report recognizing when they are "acting like someone they are not,"."being phony," "not expressing their true feelings," and "not being who you really are inside." The motives both boys and girls give for such behavior are to gain acceptance, seek approval from peers and parents, and improve social relationships. Sometimes the concern that others will not understand them is expressed (Harter, 1992). Harter found that where self-acceptance was contingent upon external approval, motivating the portrayal of a false-self, levels of self-esteem were significantly lower. To explore the relation between the various conventional care scales and a false-self, participants in the present study were asked to respond to Harter's two self-orientation questions on a 7-point Likert scale. False-self If others like or approve of me (first) then I will like and approve of myself. True-self If I (first) like and approve of myself, then others will like and approve of me. Univariate tests revealed significant gender differences on both the True-self and the False-self questions, F(l,290) = 11.40 and 5.20, ps < .001 and < .05, respectively. Men more than women believed that liking and approving of themselves preceded others liking and approving of them. In contrast, women were more inclined to believe the reverse, that self-liking and -approval was contingent upon others liking and approving of them. Zero-order correlations with care/dependency scales predicting both True- and False-self orientations are reported in Table 11. 147 Table 11 Correlations between Care Scales and Self-orientations Priority Scale True-self False-self Octant HI (Unassured-Submissive) Silence .01 .37 External -.22 .48 Social Perfectionism - Self- directed -.13 .50 Anxious Concern -.15 .49 Demand for Approval -.11 .51 Octant JK (Unassuming-Ingenuous) Care -.08 .30 Unmitigated Communion .03 .11 Perfectionistic Care .01 .23 Self-Sacrificial Care -.09 .23 Octant D E (Cold-hearted) Social Perfectionism - Other-directed -.05 .34 Conventional Care - Other-directed -.16 .38 Conventional Care - Socially Prescribed -.00 .34 Note. Correlation coefficients greater than .11 are significant at p < .05, those greater than .15 are significant at p < .01, and those greater than .18, p < .001. Care and dependency scales predicted the False-self but little relation was found predicting the True-self. Scales within Octant HI (Unassured-Submissive) were the strongest predictors of a False-self orientation. Perhaps curiously, of these, Silence proved the weakest predictor. Scales within Octant D E (Cold-hearted) were moderate predictors and scales within Octant JK (Unassuming-Ingenuous) low to moderate predictors. Gender Differences: Motives for Helping To investigate the question of motivation, participants were asked to write two narratives of care. After each, a measure of change in negative emotional state was completed along with a one-item response indicating the level of rumination the incident 148 evoked. In the first narrative participants were asked to "write a story about a time when a friend (or partner) either ignored or refused your help or care." The second narrative involved writing "a story about a time when a friend (or partner) went to someone other than yourself to get and receive assistance or care." The purpose of having participants actually write the stories was primarily to reactivate the initial emotional arousal. A summary or global distress score was calculated based on both stories and included the rumination item. Authentic care would involve less identification with one's role as a care-giver and greater contextual sensitivity regarding the friend's actually obtaining help. Refusing an individual's help need not imply a negative comment about the would be care-giver. The person in need of help may seek out a professional such as a counselor or another individual with greater expertise in the area, or they may simply want to learn for themselves. (Both of these comments were mentioned in the care narratives themselves.) Refusal to receive help may also indicate a certain insensitivity in the manner in which care is offered. Insensitive caring may entail relational fatality (Hennig, Myers, & Walker, 1999). Zero-order correlations with distress are reported in Table 12. The highest correlations with care/dependency scales again came from Octant HI and DE, and low to moderate correlations were associated with JK. Caring from a conventional perspective is less about actual empathic concern or even awareness of the other, but often serves instrumental purposes of seeking approval or fending off expectations of (perceived) negative evaluation. Batson (1995) described these two motives as the empathy-altruism hypothesis versus the empathy-specific reward hypothesis. The gullibility and difficulties in saying "no," reflective of Octants HI and JK, make it likely that caring or helping is also motivated by concerns about the other's response. Amongst all the care/dependency scales it was the Octant HI Demand For Approval Scale that showed the strongest relation to experiencing distress when a friend refuses help or seeks help elsewhere. High scorers have over-identified with the self they believe is reflected in the eye of the other, the "looking glass self." 149 Table 12 Correlations between Care Scales and Care Narratives Scale Total Men Women Octant HJ (Unassured-Submissive) Silence .19** External .31*** .15 .39*** Social Perfectionism - Self-directed .33 .16 .40 Anxious Concern .31 Demand for Approval .35 Octant JK (Unassuming-Ingenuous) Care .20*** Unmitigated Communion .31 Perfectionistic Care .23 .05 .32 Self-Sacrificial Care .26*** .06 .36*** Octant D E (Cold-hearted) Social Perfectionism - Other-directed .33*** Conventional Care - Other-directed .22*** Conventional Care - Socially Prescribed .34*** *p<.05 **p<.01 ***p<.001 Care and dependency scales predicted the False-self but little relation was found predicting the True-self. Scales within Octant HI (Unassured-Submissive) were the strongest predictors of a False-self orientation. Perhaps curiously, of these, Silence proved the weakest predictor. Scales within Octant D E (Cold-hearted) were moderate predictors and scales within Octant JK (Unassuming-Ingenuous) low to moderate predictors. For each of the pairs of nine self-directed scale correlations, comparisons across gender (using Fisher r to z transformations) were computed. Comparing correlations for each of the nine scales predicting distress revealed that women's correlations were in all instances higher but not necessarily significantly higher than men's. On Perfectionistic Care, Self-sacrificial Care, External, and Self-directed Social Perfectionism women did, however, report a significantly stronger relation to distress than did men. Where significant differences 150 were found, separate correlations are reported above in Table 12. While men report similar conventional ideals of striving to be perfectly caring and placing others' needs above their own, women as a group are significantly more distressed when their cultural role as care-givers is not reciprocated. Regression Analyses Predicting Adjustment The care literature has been largely focused on predictor variables for depression. The above investigations sought to examine gender differences in predicting depression, but found none. Demonstrating differences, however, between men versus women, depressed versus non-depressed groups, or correlations with other clinical phenomena is insufficient to warrant the existence of a scale. Authors of proposed scales need to demonstrate that these are not simply redundant with, or a proxy for, more global measures of adjustment like self-esteem. Hierarchical regression analyses were thus conducted with the adjustment measures (global self-esteem and depression) as outcome variables in an effort to determine the predictive capacity of the respective care scales over and above already well-validated octant scales from the Interpersonal Circumplex. Neuroticism was also included, given its strong correlation with any form of maladjustment generally. Hierarchical regressions were undertaken for each of the three octants (DE, HI, and JK) together with their respective care/dependency scales. While not significant in the current study, as gender differences are typically found for depression and neuroticism, the first predictor block was composed of gender. Having controlled for gender, the two Interpersonal Circumplex scales defining the particular octant were entered next. Conceived as lower-order scales within the more global circumplex octant it was important to demonstrate that the more narrowly clinical care/dependency scales could contribute beyond the more global octant scales. Given the strong relation between neuroticism and depression, it was also important to demonstrate that scales predict over and above a global measure of maladjustment, such as neuroticism. In the third predictor block, then, Neuroticism was entered, and finally the care/dependency scales reflective of the particular octant were 151 entered as a block. Details of the regression analyses can be seen in Appendix F. The general results of the regression analyses indicated that Octant FU (Unassured-Submissive) scales were a strong predictor of depression (A/?2 = .13, p < .001) and self-esteem (A/?2 = .22, p < .001), controlling for gender; whereas Octant scales D E (Cold-hearted) and JK (Unassuming-Ingenuous) contributed little further variance in the equation. The strong relationship between Neuroticism, entered in on the third step, and measures of adjustment was evidenced across the regressions for both self-esteem and depression. The specifics of each octant, relative to the additional contribution of care scales, will be taken up in turn. Beginning with Octant DE, and with all the respective variables entered into the regression equation, the Other-directed Conventional Care Scale was the only significant variable from those in the final step to predict low self-esteem (3 = .20, p < .001). Zero-order correlations for Other-directed Social Perfectionism and Conventional Care, and Socially Prescribed Conventional Care with depression (rs = .32, .31, and .28, ps < .001, respectively) and self-esteem, coded in the reverse or low direction (rs = .30, .34, and .25, ps < .001, respectively), are moderate. Recall that no scales projected into Octant FG. For Octant HI, only the Silence and Anxious Concern Scales contributed significant additional variance in predicting self-esteem (3 = .16 and .14, p < .01 and .05, respectively). Only as a block were HI care/dependency scales able to contribute any additional variance in predicting levels of depression (A/?2 = .02). Note that the External Subscale was not included in these analyses owing to the above-identified cluster of self-critical items which overlap extensively with items found in Beck's Depression Inventory (e.g., feeling like a failure, guilty, self-disappointment, self-blame) and evidenced by their very strong correlation (r = .61). Thus, the inclusion of the External Subscale would distort the relations that exist between Octant HI scales and indices of adjustment. Zero-order correlations for Self-directed Social Perfectionism, Silence, Anxious Care, and Demand for Approval with depression (rs = .40, .30, .41, and .38, ps < .001, respectively) and self-esteem (rs = .44, .41, 152 .51, and .45, ps < .001, respectively) are moderate to high. Amongst the scales falling within Octant JK, only Self-sacrificial Care contributed any significant variance in predicting low self-esteem ((3 = .17, p < .01), and no additional variance in depression was accounted for. Zero-order correlations for Care, Unmitigated Communion, Self-sacrificial Care, and Perfectionistic Care with depression (rs = .26, .23, .19, and .13, ps < .001, .001, .001, and .05, respectively) and self-esteem (rs = .24, .15, .23, and .12, ps < .001, .01, .001, and .05, respectively) are low to moderate. The results of the regression analyses indicate that many of the care scales, particularly those developed outside the present study, contributed little beyond already existing octant scales in conjunction with Neuroticism, at least insofar as they are able to predict depressive symptoms. The only scale to contribute additional significant variance in predicting depression was Other-directed Conventional Care. As a more global marker of adjustment, the scales contributing additional variance in predicting low self-esteem were Other-directed Conventional Care, Anxious Concern, Self-sacrificial Care, and Silence. Two points are worth noting. Only those scales that were demonstrated in the current study to meet the more rigorous inclusion of circular criteria as well as loading uniquely on their respective conjoint factor loadings proved predictive. These involved the scales developed in the current study, with the addition of the Silence Scale, which has consistently shown itself to be robust and distinct. It should be noted that these regression analyses serve as a strong test as they involved controlling for gender, two octant scales, and neuroticism, which in itself is a very strong predictor of both depression and self-esteem. Future research will need to examine the comparative predictive validity on an hypothesized act criterion (e.g., peer ratings). These regression analyses have thus presented a conservative case; most predictive were those scales within octants adjacent to FG, where low self-esteem and depression were projected. While scoring high on the care/dependency scales generally is associated with maladjustment, it could also be the case that scoring low might be predictive of problems. In other words, a curvilinear, or U-shaped, relationship might be evident. Thus, in addition to 153 the hierarchical regression analyses conducted between measures of care and adjustment (reported above), additional regression analyses were conducted to estimate curves (viz., linear, quadratic, cubic) for all twelve care/dependency scales, including the non-self-directed scales, in predicting depression and low self-esteem. Where the relationship is linear, data points within the scatterplot will cluster around a straight line. Where the relationship between two variables is not linear, quadratic and cubic regression models can be fitted to the data and evaluated for goodness-of-fit by examining the change in R2 for each model. The more coefficients added to the regression equation, the higher AR2 becomes; for example, when fitting a quadratic over a linear model to the data. Procedurally, the addition of coefficients stops when no significant model improvement is gained. Results for these analyses indicated that of the twelve care/dependency scales, only three (Self- and Other-directed Social Perfectionism and Anxious Concern) evidenced significant quadratic relations with depression and only the Silence scale evidenced a quadratic relation with low self-esteem (see Table F-4 in Appendix F). No cubic relationships were found. Consider the scatterplot for Anxious Concern in predicting depression (see Figure 26). While the quadratic model contributed a significant AR2 over the linear model, contributing an additional 2% to the overall model fit, an examination of the quadratic trend-line indicates that the U-shaped relationship is not particularly meaningful when compared against the accompanying linear trend-line. At the level of test interpretation this suggests that respondents are reading the low end of the 7-point Likert scale as unipolar not bipolar; strongly disagreeing that one is perfectly caring does not indicate that the respondent considers him- or her-self as Cold-hearted (Octant DE). Consistent with the interpersonal circumplex, scales are unipolar. Bipolar scales would likely load on more than one octant scale and thus reduce the circular factor loading. Recall that, within the interpersonal circumplex approach, reverse items, or in the present context, distinct halves of the U-shaped quadratic relationship, are expressive of separate octant domains. 154 Figure 26 Scatterplot of Anxious Concern and Depression, with Linear and Quadratic Trend-lines 0 10 20 30 40 50 60 anxious concern Discussion The global purpose of this study was to undertake a structural "mapping" of where care can go awry, referred to here as conventional care. The use of the qualifier "conventional" signals a motivational orientation of living up to social norms and shared expectations of what it means to be a good person for the purpose of gaining social approval. In Kohlberg's terms this is a "good/bad boy or girl" orientation. A super circumplex space was deemed a relevant domain for such an investigation. The first prescriptive use of the circumplex was to place authentic care both flexibly and with moderate intensity (i.e., vector length) within the upper-right Quadrant I of interpersonal space (see Figure 2 above). Doing so empirically operationalizes authentic care as at least involving a balance of both flexibility-rigidity and dominance-nurturance (or self-definition and self-relatedness), recalling Aristotle's Golden Mean. In contrast, conventional care was located in Quadrant IV as anticipated, but examining the projection of provisional items further suggested the 155 development of additional scales relevant to Octants F G and NO: Selfishness and Intrusive Care, respectively. Like dependency, the interpersonal aspects of conventional care proved to span a broad array of meanings, hitherto not grasped. Following on this, the second prescriptive use of the Interpersonal Circumplex was at the item level in using the geometric characteristics of circular space to suggest the addition of circular criteria to more traditional approaches to test development. The current study proposed the use of theoretically guided structural criteria in addition to the usual substantive approaches for the development of interpersonal measures. Two circumplex propositions grounded the development of such criteria: (a) the correlation between any two variables in interpersonal space is a function of their angular proximity within that space; and (b) several scales may occupy one and the same octant, sharing an underlying interpersonal style, yet possess sufficient unique variance to suggest the need of substantive analyses (i.e., item-total correlations, factor analysis) in item selection. This study represents the first of its kind to have suggested specific criteria and actually employed them in the construction of scales. The validity of the approach was supported in the present investigation of existing care/dependency scales where it was demonstrated that an item's contribution to the parent scale's reliability was directly proportional to its (absolute) deviation from the circular mean. A thematic investigation of items supported the picture drawn from alpha-contribution plots as well as the item conjoint principal components analysis; thus integrating circular structure and substantive approaches. In total, three self-directed scales and two non-self-directed scales were developed, extending the mapping process into additional interpersonal dimensions, other-directed and socially prescribed conventional care. It has been noted throughout that scales and items within the self-directed dimension fell across more than one octant, suggesting distinct structural and thematic differences. In this, the care domain like the dependency domain is represented by an overly broad array of semantic meanings from which existing scales too carelessly sample. A conjoint principal 156 components analysis confirmed the frequent existence of more than one cluster of items within the same scale. Jack's External Scale possessed three distinct clusters and the Self-directed Social Perfectionism Scale possessed at least two. Other distinctions within one and the same scale were noted. Together, the items were represented by six distinct factors. Four factors (i.e., Anxious Self-critical Concern, False-self, Negative Assertion-Reverse, and Demand for Approval) fell within Octant FU (Unassured-Submissive) and share what Wiggins (1995) has called a "masochistic" orientation involving granting neither love nor status to the self, and status but not love to the other. Two additional factors (i.e., Conventions of "Goodness" and Perfectionistic Care) fell within Octant JK (Unassuming-Ingenuous) sharing a motivational orientation Wiggins (1995) labels "martyrdom." The scale, along with the factor, are labelled "Anxious Concern" rather than "Anxious Care" because of the ambiguity about the object of concern, the self or the other. Their location in Octant FU would suggest that concern largely involves defending a self deemed unworthy of an authentic concern, which ultimately is given neither to the self nor the other. Octant JK shares with FU the withholding of love and status to the self, but unlike Octant Ffl, both love and status are granted to the other. Scales falling within the JK octant are reflective of a second cluster of themes found in the care literature involving self-sacrificial care for others and conventions of goodness understood as attributing higher value to the needs of others than the self. But rather than aggregate the scale scores together as users of the Silencing the Self Scale have done, it is proposed that these two octants, JK and Ffl, represent two meaningfully distinct groups of scales which are overlooked otherwise. The importance of anchoring scales, particularly newly developed ones, within a well validated nomological net is of importance to constrain the range of interpretive hay that is made out of tests which fail to even match their titles. For example, the Self-directed Social Perfectionism Scale had less to do with realizing perfect social relationships than it did with being socially anxious and fearful. Two points need to be conjoined. First, scales within the 157 same octant were, for the most part, more closely associated with one another than they were with scales from the neighboring octant. Second, the circular projection of circumplex scales suggests that the individual profiles will possess a peak elevation in one of the eight octants, with lesser elevations in adjacent octants. The two points combined suggest that a valid distinction can be made between two types of conventionally caring/dependent personalities, or "characters:" one projecting into Octant HI and the other into Octant JK. Interpersonal theory would suggest that the bids that individuals make in the intersubjective construction/construal of the self tend to be self-confirming. For example, individuals with rigid and intense peak elevations on Octant HI will likely employ an anxious avoidant orientation, worrying that they will "say something tactless" or "do something inconsiderate." The implicit moral norms of care arid concern, of secondary moral self-evaluations regarding feeling "sad" and "bad," are apparent. From an interpersonal perspective such bids tend to elicit from others a self-confident domineering stance (Octant PA). Gilligan records a 19-year-old girl's concern with hurting others compared with her boyfriend's more straightforward approach. I never want to hurt anyone, and I tell them in a very nice way, and I have respect for their own opinions, and they do things the way that they want. He usually tells people right off the bat. He does a lot of things out in public which I do in private. It is better, but I just could never do it. (1982, p. 80) Jack (1991) and others (e.g., Thompson, 1995) suggest that conventional care is enacted "for the sake of maintaining relationships," where this means compulsive care-giving and anxious attachment. From an interpersonal perspective, compulsive care-giving, however, is an Octant NO construct and I have gone some distance in the current study to confirm this. Previous research indicates that anxious attachment is an Octant L M construct (Bartholomew & Horowitz, 1991). In contrast to both compulsive care-giving and anxious attachment, the current study indicates that conventional care has two faces: Octant HI and Octant JK. Scales within Octant HI share more with fearful attachment than they do with anxious attachment (Bartholomew & Horowitz, 1991). Octant JK shares features of both 158 anxious and fearful attachment, but from an interpersonal perspective is distinct from both. Unassured-submissive individuals will silence themselves and evoke domination. The interpersonal complement of Octant HI modest self-effacing behavior is Octant PA demanding respect, and vice versa (Carson, 1969). The only way to explore relationships is by avoiding social blunders, rather than by fostering the more positively oriented interpersonal skills of self-disclosure, negative assertion, and genuine emotional support. Harter (1986) has been pursuing the question of why girls' levels of self-esteem decline after Grade 3, whereas self-esteem in boys remains relatively stable. Women, particularly conventional women in the role of homemaker, have been found to report lower levels of self-esteem than men (Harter, 1992). Recall that Gilligan's theory holds that young girls repress and dissociate from their "true" selves to take up the voice of the detached other (i.e., patriarchal conventions of feminine goodness as self-sacrifice). Women's true self includes the value of relationships, caring, and cooperation rather than detached competition. Concurrent with the decline of self-esteem amongst girls in early adolescence is the developmental emergence of a false-self. Individuals report recognizing when they are "acting like someone they are not," "being phony," "not expressing their true feelings," and "not being who they really are inside." The motives both boys and girls give for such behavior are to gain acceptance, seek approval from peers and parents, and improve social relationships. Sometimes the concern that others will not understand them is expressed (Harter, 1992). Harter found that where self-acceptance was contingent upon external approval, motivating the portrayal of a false-self, levels of self-esteem were significantly lower. By taking up an external false-self orientation, judging oneself by how others see one, such individuals place themselves in a double bind. As relationships inevitable chaff and cause strife, the approval is all the more needed but anger must also be held inward. As is usually the case, more covert forms of aggression, like walled-off silence, withdrawal, and slamming doors, will only increase the alienation, relational distance, and rejection. 159 Behaviors expressive of Octant EQ are far more reflective of Gilligan's Level 1 care. Level 1 care is primarily a strategy for coping with hurtful rejection and domination; seeking "survival" through self-protection (Gilligan, 1982). "Women in some instances deliberately choose isolation to protect themselves against hurt" (Gilligan, 1982, p. 75). The "sense of isolation, aloneness, powerlessness," as Gilligan puts it, can often be the impetus to move to the neighbor Octant JK Level 2 care, where the self is neglected but at least one is in a position of less isolation. Reflective of the experience of being an adopted child, a young girl comments at Level 1 with regard to the Kohlbergian Heinz dilemma. The druggist is ripping him off and his wife is dying, so the druggist deserves to be ripped off. (Is it the right thing to do?) Probably. I think survival is one of the first things in life that people fight for. (Gilligan, 1982, p. 110, italics in original) Individuals projecting into Octant JK (Unassuming-Ingenuous) possess greater interpersonal skill. While continuing to withhold caring and status from the self, they do grant both to others. Scales within Octant JK share a concern with others, denying self but affirming the importance of others' needs. While gender differences were not found on any of the self-directed care/dependency scales, the correlations between two of the JK scales (i.e., Perfectionistic Care and Self-sacrificial Care) and negative arousal/rumination in the care narratives were significant for women but not for men. External and Self-directed Social Perfectionism also were predictive for women, but it was hypothesized that this arose largely from their JK clusters; further illustrating the interpretive problems that accompany overly broad scales. Self-identifying with the care-giving role has been here characterized as a largely Octant JK phenomenon. While men and women can be said to strive equally for these values, it would appear actually doing so or identifying with such values is more central for women than it is for men. Increasingly gender differences have been explored not as main effects, but in their differing relations to other measures and processes. One direction examining the increased frequency, severity, and duration of depressive episodes for women, has focused on differing 160 coping strategies. Women are reported to enlist social support for emotional and instrumental reasons, including venting of emotions, whereas men report greater alcohol use in stressful situations (Carver, Schier, & Weintraub, 1989). Similarly, Nolen-Hoeksema and colleagues found that women engaged in a more "ruminative" response style to depressive moods than did men (Nolen-Hoeksema, Morrow, & Fredrickson, 1993). The picture of women as engaged in a "less adequate" emotional coping style versus men's more adaptive problem-solving style has not been without controversy, however (e.g., Banyard & Graham-Berman, 1993). Excessive focus upon others has its costs, not only to the self but to the actual quality of the care engaged. Alienated from oneself, not being aware of one's needs and desires is to be cut off from those emotions and affects which constitute the grounds of authentic caring and morality. Indeed, the view from the looking glass self, the view of the other, is the view of the self. Taking the looking glass metaphor in its most literal form (Harter, 1992) involves a heightened concern regarding one's physical bodily presence and a self-consciousness of its desirability in public space. Comparisons with Bruch's (1973, 1988) seminal descriptions of eating disordered women and their "not owning their own bodies, as not having a center of gravity within themselves.... actfing] as if their body and behavior were the product of other people's influences and actions" (1973, p. 55) come ready to hand as extreme examples. The external orientation is well captured in Bruch's experience of eating disordered women continually "double-tracking," in that they are always self-consciously monitoring what they are doing in light of the contextual cues of others. In the next study the relation of conventional care to eating disordered behaviors was examined. 161 CHAPTER 3: CONVENTIONAL CARE AND EATING DISORDERS Introduction Gilligan's move, in what came to be called the Kohlberg-Gilligan debate, was described as one of linking an interdependent account of selfhood with gender and an ethic of care. Across several studies reported in Chapter 1, women were reported as possessing a more interdependent, less bounded, sense of self than men, in support of Gilligan's (1982) thesis. Within the domain of moral psychology, research has focused on the two "voices" of justice versus care, investigated largely with respect to gender. If a distinct conception of morality can be identified in so large a category as that constituted by women, then Kohlberg's "mapping" of the moral domain has indeed been overly narrow in scope. Empirical investigations reveal that women do raise more care concerns, coexistent with the greater number of personal dilemmas they raise, reflective of the continued role (and related expectations) that women play as nurturers and maintainers of social relations. Consistent with Gilligan's stance, her ethic of care was less a thorough-going philosophical theory than it was an applied concern for the struggles of women. While Gilligan and her followers have largely focused on depression (e.g., Steiner-Adair, 1990), eating disorders represent an even stronger and more potentially fatal concern. Jack's (1991) Silencing the Self Subscales have become the first broadly employed self-report measure of Gilligan's ethic of care; reflecting themes of repressing one's opinions where they may cause disharmony or disagreement (i.e., Silence Subscale), beliefs and attitudes regarding the greater importance of others' needs, the desire to make others happy first, care as self-sacrifice (i.e., Care Subscale), and defining oneself from the perspective of significant others (External Self-perception Subscale). Jack also includes a fourth Divided-self Subscale that was not studied in this dissertation, owing to its overlap with several currently existing scales of its kind (e.g., Campbell, 1990), and its distinctiveness from core ethic-of-care notions. While the relation between self-silencing and depression has been studied, apart from the current data set, their relation to eating disorders has not been investigated. 162 The purpose of the current investigation was to examine the construct of conventional care, for which there exists only the current data set amongst a largely clinical population, against the construct of perfectionism, for which there exists an extensive literature in the prediction of eating disordered attitudes and cognitions. Specifically, conventional care (i.e., Jack's "self-silencing") will be compared with perfectionism in the prediction of (a) adjustment and (b) weight/shape concerns. While the current secondary analyses find their place at the end of this dissertation for illustrative purposes indicating future clinical directions, it is worth noting that the promising earlier results from this data set (Geller, Cockell, & Goldner, in press) which included Jack's (1991) Silencing the Self Scale, are what first precipitated the need for their clarification. Given the lack of psychometric procedure in developing the instrument, it remained unclear precisely what the scales were in fact measuring. The success of the present investigation with its more differentiated mapping of the ethic-of-care domain owes its initial impetus to the insights operationalized within Jack's scales. The initial study examined differences among three groups of women matched on education: an anorexic group, a general psychiatric group, and a normal comparison group. Individuals not meeting all DSM-IV criteria for anorexia nervosa were excluded from the first group. Individuals within the second and third groups, the two comparison groups, who met DSM-IV eating disorder criteria, either currently or in the past, were excluded from their respective groups. Geller et al. (in press) found that the anorexic group reported significantly greater scores on all four STSS Care, Silence, External Self-perception, and Divided-self Subscales (Jack, 1991), as well as on Spielberger et al.'s (1985) Anger-in Scale. (Recall that a small provisional set of anger-in items was identified as falling into Octant HI; along with Silence, External, Demand for Approval, and. Anxious Concern.) After covarying for depression, self-esteem, and global assessment of functioning - all three discriminating among groups (although there were no differences between the comparison groups on depression) - Care and Silence Subscales continued to distinguish between anorexic and 163 comparison groups. While the previous conjoint principal components analyses, alpha-contribution plots, item circular projections, and thematic contents all confirmed the existence of various clusters within scales falling largely into Octants HI and JK, the Silencing the Self Scale will serve as an adequate proxy for the then non-existent Conventional Care Scales. The distribution of Care items across both octants would also make separate octant analyses difficult to justify. Additionally, the Silence Scale proved to have very good psychometric characteristics and will be carried over into future investigations. Conventional care was thus assessed as a "global self-silencing" construct employing the same three scales used in Study 1 (Care, Silence, and External Subscales), aggregated to produce a total score. Structural equation modelling was used to investigate the role of each, global self-silencing and perfectionism. The current structural equation analyses were intended as a window into subsequent and more extensive clinical studies. This needs to be stated, lest the brevity of the final analyses be seen as insufficiently warranting the extensive introduction that follows. By way of a brief overview then, first, the epidemiology and characteristics of eating disorders will be discussed, followed by a review of pertinent literature bearing on two potential explanatory areas: the socio-cultural emergence of the thinness ideal and developmentally related aspects of identity formation. Then the constructs of care and dependency will be discussed in light of the little research that has been undertaken amongst eating disordered women. The second key construct of concern, in addition to conventional care, perfectionism, will then be discussed prior to an explication of the structural equation models. Eating Disorders: Epidemiology and Characteristics The prevalence of eating disorders ranges from 1% to 4% of adolescent and young women from predominantly upper-middle and middle social classes. While the incidence is lower elsewhere in the population (Szmukler, 1985), rates seem to be increasing among 164 males, minorities, and women of all ages (APA Practice Guideline for Eating Disorders, 1993; Crisp, Palmer, & Kalucy, 1976). Developmentally, only a few cases appear among prepubertal children, although recent trends have shown an increase in this age category. At the ages of 11 and 12 rates of onset rise steeply, plateauing between 13 and 19 years of age, gradually diminishing from 20 to 30, and rare beyond 30 years of age (Theander, 1996). Groups at particular risk are dancers, models, competitive athletes, and those under strong pressure to achieve (e.g., college students). Homosexual men appear at greater risk than heterosexual men (Whitaker, Johnson, Shaffer, Rapoport, Kalikow, Walsh, Davies, Braiman, & Dolinsky, 1990). The risk increases for women with certain psychiatric disorders (e.g., depression, borderline personality disorder), life experiences (early sexual abuse), and family histories (depression, eating disorder, alcoholism) (APA Practice Guideline for Eating Disorders, 1993; Fairburn et al., 1987), although the relationship between these variables and eating disordered behaviors is complex. Mortality estimates range up to approximately 18%, but probably a more accurate estimate would put it around 5% for those with anorexia nervosa (Garfinkel & Garner, 1982). Bulimia nervosa is considerably more common than its anorexic counterpart (Whitaker et al., 1990). A considerable amount of research and theory has elucidated the psychopathology of eating disorders, generally understood to involve a variety of psychological, biological, and familial, and cultural correlates (Garfinkel & Garner, 1982). These several levels of investigation can be understood as reflecting both specific and general components. The specific components are those peculiar to anorexia and bulimia nervosa themselves, and are comprised of disturbances in eating behavior and attitudes to food, eating, weight, and shape. Considerable consensus exists regarding the specific components of eating disorders as they are set out in the current Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; American Psychiatric Association, 1993). The physical criteria for anorexia nervosa include a "refusal to maintain bodily weight at or above a minimally normal weight for age and height" and amenorrhea, or loss of menses. Cognitive components include an "intense fear 165 of gaining weight or becoming fat" even though the individual may be underweight, and one or more of the following: "disturbance in the way in which one's body weight or shape is experienced" (e.g., feelings of fatness, avoiding exposure); "undue influence of body shape and weight on self-evaluation," and/or "denial of the seriousness of the current low body weight" (see Appendix G for DSM-IV eating disorder criteria; Cooper & Fairburn, 1993). Although extreme weight- and shape-preoccupation represents the shared core, the principal difference between eating disorder subtypes is the presence of low weight in the restricting and bulimic subtypes of anorexia nervosa, and the binge-purge cycle in bulimic anorexia nervosa and normal-weight bulimia. The more general components include those aspects of impaired psychological functioning which eating disorders share with other psychiatric categories. Eating disorders are often found to be comorbid with a variety of psychological symptoms. For example, using DSM-III-R criteria, Wonderlich, Swift, Slotnick, and Goodman (1990) found a variety of personality disorders in their sample of eating-disordered women: dependent (DPD; 32%), avoidant (APD; 32%), borderline (BPD; 25%), histrionic (HPD; 23%), and obsessive-compulsive (OCD; 18%). Different investigators examining diverse samples reveal a similar range of psychopathology (for a review, see Vitousek & Manke, 1994). Comorbid major depression and/or dysthymia have been reported in 50-75% of anorexics (Halmi, Eckert, Marchi, Sampugnaro, Apple, & Cohen, 1991). The classic anorexic displays a familiar and coherent cluster of rigid and restrained characteristics. She is a tense, hyperactive, alert, rigid person. Usually she walks, talks, and thinks rapidly. She is inordinately ambitious, drives herself hard, is markedly sensitive, and obviously feels insecure. An immature and severe conscience guides her actions and she is said to be hyperconscientious. Neatness, meticulosity, and a mulish stubbornness not amenable to reason make her a rank perfectionist. Usually she is introverted, serious, self-willed, and lacking in the warmth and spontaneity that are consistent with her years. (DuBois, 1949, cited in Vitousek & Manke, 1994, p. 137) Bulimics are in many respects diametrically opposite, impulsive in characterization, 166 manifesting an erratic consummatory pattern marked by the alternation of restraint and disinhibition. The binge-purge cycle is thought to serve the functional purpose of affect regulation (Johnson, Lewis, & Hagman, 1984) and temporary suppression of painful self-awareness (Heatherton & Baumeister, 1991). Many bulimic clients report dissociative symptoms, sexual problems, and a variety of impulsive behaviors such as overspending, shoplifting, promiscuity, and self-mutilation (Favazza, DeRosear, & Conterio, 1989; Wilson, Hogan, & Mintz, 1985). The assumption that some dissimilar premorbid traits must be responsible for the divergence of anorexic and bulimic symptoms in otherwise similarly weight- and shape-preoccupied individuals has propelled no small amount of research. In general, the question of whether the eating disorders should be found expressive of some discrete cause has been highly controversial, caught up in questions regarding what is "core" to the disturbance (Cooper & Fairburn, 1993), the discontinuity of eating disorders with dieting behavior (Polivy & Herman, 1987; Vredenburg, Flett, & Krames, 1993), and medical nosology generally. Attempts at seeking after some single discrete cause of eating disorders is generally regarded as wrongheaded (Hsu, 1990), in favor of a multidimensional conceptualization (Garfinkel & Garner, 1982). For example in the field of medicine: hypertension may be due to coarctation of the aorta, excess renin or aldosterone, or a pheochromocytoma (Weiner, 1977). Hypertension, like schizophrenia or eating disorders, is best conceived as a syndrome with no single common pathogen. Developmental psychopathology's notion of "equafinality" (i.e., a number of distinct developmental pathways finding a common terminus) is likely the more veridical account (Cicchetti, 1990); up to 50% of those with anorexia nervosa will develop bulimic symptoms, significant numbers of individuals who are initially bulimic will manifest anorexic symptoms, and restricting and bulimic subtypes may occasionally alternate within the same client (Garfinkel, Moldofsky, & Garner, 1980). Historically, many investigators are of the opinion that the prevalence of eating 167 disorders has been increasing over the past two or three decades (Kendler, MacLean, Neale, Kessler, Health, & Eaves, 1991; Lucas, Beard, O'Fallon, & Kurlan, 1991; Russell, 1995), although this may be more true of bulimia (Brownell, 1991; Cooper, Charnock, & Taylor, 1987). Estimates are complicated by differences in research methodology among the 50 some epidemiological studies thus far conducted. A clear estimate of development is complicated by a number of historical factors. An increased tendency to seek treatment, along with a greater likelihood of being diagnosed and referred, might be coincident with the increased publicity eating disorders have recently received. The question of central interest remains: Why is the thinness ideal prized primarily among women in our culture? Why now? Why young women in particular? The Thinness Ideal: Why Women? Why Now? Compromising physical health and comfort for the sake of cultural ideals is not new. In pre-revolutionary China the thousand-year-old custom of footbinding resulted in a clubbed "lily-foot" so greatly admired in its signification of high social standing. The wearing of corsets in the nineteenth century, while signifying beauty and purity, caused extreme discomfort and digestive problems. Sontag (1978) documents how the wasting illness of tuberculosis had become glamorized in the nineteenth century as "one index of being genteel, delicate, sensitive. It became rude to eat heartily. It was glamorous to look sickly" (p. 28). "Consumption" (i.e., especially pulminory tuberculosis) was the initial category from which anorexia nervosa was first distinguished ("nervous consumption;" Garfinkel & Garner, 1982). The male preference was for women who were delicate and pale. Women would use whitening powder rather than rouge and drink lemon juice and vinegar to dampen their appetites to maintain a pale appearance (Vincent, 1979). Everything from tattooing, scarification, cranial or bone distortion, to ornamental tooth fillings have been socially prescribed - many primarily upon women (Blinder & Chao, 1994). The thinness ideal seems the most current form of idealized beauty held out for women today. The Rubenesque nude with her dimpled and abundant flesh is a classic 168 example of the endomorphy preferred in previous eras. Our current century has gone from the buxom woman at its turn, to the flat-chested "flappers" of the 20s, back to the "full-figured" Marilyn Munroe of the 50s and 60s, and since then, through Twiggy in the mid-60s to the inclusion of today's more muscularly lean look. Researchers conducting content analyses have reported a dramatic increase in the number of articles on diet, exercise, and diet/exercise in six of the most popular women's magazines between 1959 and 1988 (Wiseman, Gray, Mosimann, & Ahrens, 1992). While the level of dieting articles seems to have leveled off in 1981, the number of exercise articles continues to climb. Since the study conducted by Garner, Garfinkel, Schwartz, and Thompson (1980), Wiseman et al. (1992) have reported that the weight of women in Playboy centerfolds seems to have remained extremely low at 13% - 19% below that expected, and the weight, ratio of body measurements, and hip size of Miss America contestants has continued to decrease. Note that inclusion criteria for anorexia nervosa include a current weight of 15% below expected weight; so exemplarity in women is now in the diagnosable range. Wiseman et al. also report the increase in women's muscle magazines and body-building. So powerful is the thinness ideal in Western culture that the societal preoccupation with dieting and weight loss has become virtually normative, both descriptively and prescriptively, especially among women. Concurrent with these shifts has been the "almost 'epidemic rise' in the incidence of eating disorders" (Wiseman et al., 1992, p. 89; Brisman & Siegal, 1984; Garner et al., 1980). Like eating disorders (and it might be added, conventional notions of goodness as self-sacrifice), the thinness ideal occupies the same white upper-middle and middle class niche, although both seem to be currently spreading across social classes and cultures (Schmidt, 1993). The Jewish "zaftig" woman or African American large hipped and buttocked, pear-shaped woman now connotes the peasant, the woman of the old world. Since even a decade ago, the large minority woman, more acceptable in the African American community with its stereotype of the desexualized "Mammy," has given way to the Madonna look. Over the past 169 decade Essence, a magazine dedicated to Black women, has increased its coverage of ever changing diet fads, exercise regimens, and dieting success stories (Bordo, 1993). The inverse relationship between social class and weight is diminishing across cultures (Dornbusch, Carlsmith, Duncan, Gross, Martin, Ritter, & Siegel-Gorelick, 1984; Stunkard, D'Aquili, Fox, & Filion, 1972). Recent cross-cultural work by Lee (1993) has provided evidence that as Western standards for female beauty are rapidly being adopted in other cultures, the thinness ideal is being imported with them. Chinese women in Hong Kong increasingly report feeling fat in their stomach, thighs, hips, and waist. Similar to the West, body dissatisfaction is predictive of dieting behaviors. Men ... are likely to view long, slim legs, a flat stomach, and a firm rear end as essentials of female beauty. Unmuscled heft is no longer acceptable.... Even Miss Soviet Union has become lean and tight, and the robust, earthy actresses who used to star in Russian films have been replaced by slender, Westernized types. (Bordo, 1993, pp. 102-103) The result for women who cannot achieve such unrealistic standards of thinness is a great deal of "body hate." "You know perfectly well we hate our bodies," says Rachel, who calls herself the pig. She grabs the flesh of her stomach between her hands. "Who could love this?" (Chernin, 1981, p. 25) It may seem odd that an internalized judgment regarding an apparently singular aspect of our totality as persons may be so shamefully devastating to the whole. Questions as to why the thinness ideal is prized in Western culture usually evoke some historical discussion around the shifting values and cultural meanings ascribed to different sorts of physiques or bodies. The Twiggy who emerged in the 60s may have become suffused with connotations of death as a result of the spread of AIDS with its figure of sickly emaciation. Jamie Lee Curtis and Madonna are harder, leaner, more vascular and muscular in the late 80s and 90s. The thin body has become increasingly marked with health, power, and surgency and is a much better signifier of control than were earlier constructions. The attendant connotations of our culture's thinness ideal most often mentioned are of 170 two sorts: the first focuses on the aesthetics of beauty, associating sexual attractiveness with the ideal body; the second emphasizes less immediately intrinsic qualities attached to various bodies within a nomological cultural web of meanings involving social class, health, and other contemporary values. Primary among the latter group are behaviors inferred from the slender body itself, with thinness indicating a high degree of self-control, for example, that is presumed to be required to maintain the ideal. Self-control is infused with certain notions of self-discipline as well as romantic notions of self-construction and the plasticity of selfhood. Here cultural ideals and physiology collide (Brownell, 1991). Naomi Wolf, author of The Beauty Myth (1991), has perhaps been most popularly outspoken on the first of these. The beauty myth tells a story: The quality called "beauty" objectively exists. Women must want to embody it and men must want to possess women who embody it. This embodiment is an imperative for women and not for men, which situation is natural because it is biological, sexual, and evolutionary: Strong men battle for beautiful women, and beautiful women are more reproductively successful. Women's beauty must correlate to their fertility, (p. 12) Thinness has become associated with attractiveness, which in turn is associated with a host of other constructs. Investigators report an asymmetry in favor of the body over the face in what constitutes "attractiveness" for women. A woman whose face is very attractive but whose body is not, is rated low in overall attractiveness; when her body is attractive and her face is not, overall attractiveness remains fairly high (Alicke, Smith, & Klotz, 1986). Attractiveness is not without its negative connotations. While attractive women are judged to be more vain, materialistic, and less relationally faithful (Cash & Duncan, 1984), the balance seems to be in favor of attractiveness as a positive attribute. Attractiveness is assumed to predict status and success, awarded not on the basis of earned merit but simply on the basis of appearance (Kalick, 1988). There would seem to be greater cross-cultural convergence on what constitutes an attractive face than an attractive body. The evidence suggests that males respond chiefly to 171 two facial types. The first facial type describes more childlike features, including large widely spaced eyes, a small nose, and a small chin. Men selecting the second type describe mature features such as prominent cheekbones, narrow cheeks, high eyebrows, large pupils, and large smile. Both types were conceived as attractive for Black, Oriental, and White females (Donovan, Hi l l , & Jankowiak, 1989). Interestingly, studies show considerably greater agreement regarding the defining features of attractiveness for women than for men (Donovan et al., 1989). Significantly more so for women than men, body image is related to psychological health (Robinson, Bacon, & O'Reilly, 1993), satisfying relationships (Koestner & Wheeler, 1988), and respective femininity/masculinity (Deutsch, Kroll , Weible, Letourneau, & Goss, 1988). Across a variety of print and art media, the portrayal of men emphasizes the head and face. Women's representations emphasize their bodies. Archer, Iritani, Kimes, and Barrios (1983) refer to this as a "face-ism" bias in contrast to the "body-ism" bias representative of women (Unger & Crawford, 1996). Attractiveness, particularly with respect to the body, is generally a more salient construct for women than men. Women are more likely to be perceived as unattractive, generally, than are men (Speadbury & Reeves, 1991). The "kernel of truth" in stereotypes - in this context that a particular kind of appearance or body implies possession of certain other characteristics - is regarded as more veridical for women than men. Tseelon comments, "the woman is in a no-win situation. She is expected to embody a 'timeless' cultural phantasy, but she is not really naturally more attractive than the man" (1995, p. 79). Among the second class of associations - those inferred from the thin body in contrast to an "intrinsic" or direct marker of beauty - the one probably least significant today involves the body as a marker for social stratification. During an earlier historical period where social class was considerably more stratified and food less abundant, endomorphy was presumably indicative of an abundance of both; good quality food was a visible sign of one's wealth, power, and heritage. 172 Perhaps receiving the most attention has been the association of thinness with the ascetic and moral ideal of self-control. The ideal of self-mastery goes back at least to Plato's Republic and his conception of virtue as the rule of reason over the other (lower) components of the soul (Plato, trans. 1976, p. 430). Compare the comments of an eating disordered woman: Well, I had the willpower, I could train for competition, and I could turn down food any time. I remember feeling like I was on a constant high. And the pain? Sure, there was pain. It was incredible.... I was fighting a battle. (Sacker & Zimmer, 1987, cited in Bordo, 1993, pp. 196, 198) Susan Bordo has documented the multiple overlays of meaning and metaphor involved in media representations of gender. She writes that the movies Flashdance and Vision Quest, render the hero's and heroine's commitment, will and spiritual integrity through the metaphors of weight-loss, exercise, and tolerance of and ability to conquer physical pain and exhaustion. (In the Vision Quest, for example, the audience is encouraged to admire the young wrestler's perseverance when he ignores the fainting spells and nosebleeds caused by his rigorous training and dieting.) (1993, p. 196) "Control" has become employed as a common trope in advertisements for products as far afield as mascara ("Perfect Pen Eyeliner. Puts you in control. And isn't that nice for a change?") to cat-box deodorant ("Control. I strive for it. My cat achieves it.") An ad for Speedo tennis shoes reads, "Don't just serve. Rule." The notion of "control" is a multifaceted one. Controlling food intake may be one immediate domain where success might be perceived as more assured. Rezek and Leary (1991) found that participants who were exposed to conditions of low perceived control reactively "displaced" their desire for control onto the narrower sphere of dietary restraint. Failure in the broader social sphere required some compensation in the (apparently) more easily controllable domain of food and diet. What seems important for those participants reporting a high drive for thinness was being in control; control over one's body presumably being within one's unique grasp. Controlling diet may be viewed as a compensation for an actual or perceived sense of ineffectiveness in other aspects of one's life. But the higher the 173 standard for control, the greater the likelihood of experiencing failure and guilt. The attempt to allay guilt and shame by increasing efforts at further control, either through strict regimens or very detailed eating schedules, is likely only to produce a downward spiral. Bruch (1973, 1988), one of the theoretical pioneers in the eating disorders field, believes that the core predisposition of anorexia nervosa involves a fundamental ego deficit in autonomy and mastery over one's body, which she hears clients describe as a global sense of "personal ineffectiveness." She has provided the field with vivid descriptions of what she means by this. [Anorexic clients] experience themselves as not being in control of their behavior, needs, and impulses, as not owning their own bodies, as not having a center of gravity within themselves. Instead, they feel under the influence and direction of external forces. They act as if their body and behavior were the product of other people's influences and actions. (1973, p. 55) Related to the need for control and high self-critical standards, perfectionism has been frequently described as a central feature of eating disorders (e.g., Bauer & Anderson, 1989; Bruch, 1973; Garner, 1986; Halmi, Goldberg, Eckert, Casper, & Davis, 1977). As a personality trait, perfectionism may play an important pathogenic role in the development of these disorders. In addition to striving towards unrealistic standards for attractiveness and thinness, these women may erect similarly unattainable standards for themselves in a variety of domains. Hewitt, Flett, and Ediger (1995) found that perfectionistic strivings for oneself predicted eating disordered attitudes and behaviors. The perception that others (e.g., parents) expected one to be perfect showed similar relations in addition to predicting bulimic behaviors and body image disturbances. A number of researchers have confirmed the strong relation between perfectionistic strivings and eating disordered behaviors (e.g., Formea & Burns, 1995; Garner, Olmstead, Polivy, & Garfinkel, 1984; Thiel, Broocks, Ohlmeier, Jacoby, & SchuBler, 1995). It has been postulated for over 50 years that anorexia nervosa may be related to obsessive-compulsive disorder. The obsessional nature of anorexia - compulsive calorie 174 counting, preoccupation with the body, incessant rumination about food - is obvious. In addition, over 50% of anorexics report fears of not saying things right; fear of doing embarrassing things; need for symmetry, exactness, order; need to know and remember; and fears that something terrible will happen (Hsu, Kaye, & Weltzin, 1993). Outcome studies find a persistence of many of these problems even after long-term recovery from the specific features of anorexia nervosa (Srinivasagam, Kaye, Plotnicov, Greeno, Weltzin, & Rao, 1995). Whatever the response may be to the questions, Why women? Why now? - whether thinness is a reflection of social status or eternal beauty, or a "reactive displacement" of a perceived sense of ineffectiveness and failure in other spheres, including identity formation, it is clear that the polyhedron of understanding which encompasses the thinness ideal is a historically shifting schematic associated with a host of meanings around success, health, moral goodness, sexuality, self-management, agency, and control. The third of our questions, Why adolescence? further qualifies the previous two questions and speaks to the additional necessity of a developmental perspective necessary to address the identity issue. Developmental Aspects: Why Young Women? It has already been mentioned that plotting the frequencies of eating disordered behaviors on a time line produces a spike rising sharply at the ages of 11 and 12, plateauing between 13 and 19 years, and then gradually diminishing from 20 to 30 years of age. Currently, in North America by the age of 13, 80% of girls and 10% of boys have already begun a weight-loss diet (Melin, Irwin, & Scully, 1992). A study conducted amongst schoolgirls in London found that, while a third of their sample of dieters had stopped dieting at a 1-year follow-up, 38% continued and 20% progressed to an eating disorder, in comparison with only 3% of the nondieters (Patton, Johnson-Sabine, Wood, Mann, & Wakeling, 1990). One longitudinal study found the most consistent single predictor of eating disorder was the presence of weight concerns (Killen, Taylor, Hayward, Wilson, Haydel, Hammer, Simmonds, Robinson, Litt, Varidy, & Kramer, 1994). In a large sample of girls 175 aged 15 to 16 years, investigators (Button, Loan, Davies, & Sonuga-Barke, 1997) found that 56% reported feeling too fat and had already engaged in some form of weight control strategy. Killen and colleagues (Killen, Taylor, Telch, Saylor, Maron, & Robinson, 1987) found 11% of 15-year-old girls vomited to control weight, 8% used diet pills, and 7% abused laxatives. Thirty-two percent reported above threshold anxiety levels and 43% were below criterion for low self-esteem. Eating pathology was correlated with higher levels of self-dissatisfaction both with respect to their physical appearance and global self-esteem. Negotiating this developmental period involves a number of biological and psychosocial factors. Girls' sexual maturation is accompanied by a rapid increase in subcutaneous fat measurable by skin-fold thickness (Young, Sipin, & Roe, 1968). For adolescent girls this "fat spurt" represents one of the most dramatic physical changes connected with puberty, increasing the level of body fat some 11 kg on average (Brooks-Gunn & Warren, 1985), and being associated with desires to be thinner (Dornbusch et al., 1984). A number of investigators have proposed that anorexia may be an attempt at controlling premorbid obesity. Some studies have reported that anorexic clients weighed more at birth than their siblings (Crisp, 1965, 1970; Halmi, 1974). Nearly half of Kay and Leigh's (1954) clients reported "weight deviation of major or minor difficulty connected with eating or digestion" and Halmi (1974) reported 31% were premorbidly overweight. Other investigators report similar findings (Garfinkel & Garner, 1982), although many rely on retrospective data. Higher birth-weight and premorbid obesity may be significant predisposing factors for some individuals. In one of only a few prospective studies investigating eating disordered attitudes and behaviors, Attie and Brooks-Gunn (1989) found that body fat, negative body image, and psychopathology, but not family relations, predicted problem eating 2 years later. Normal pubertal changes like breast and hip development may be associated with efforts to control food intake. Crisp (1965, 1978) argues that anorexics are actually afraid of 176 normal weight rather than being overweight; anorexia is a "regression" from normal body proportions associated with puberty and psychosocial demands. Early maturers are more likely to have eating problems but early maturers are also more likely to be heavier in weight (Faust, 1983). Some clients, though not all, do experience early puberty, which may exacerbate psychosocial expectations of a young woman who is not yet otherwise prepared. Halmi and colleagues found a bimodal clustering of eating problems at 14 and 18 years of age (Halmi, Casper, Eckert, Goldberg, & Davis, 1979). They interpreted these findings in light of the socio-developmental challenge to dependence upon family at this time; entering large, demanding, and often impersonal high schools at 14 years of age. At 18 years of age, a further expectation of independence is placed on the adolescent with entrance into the broader world. Discussed in Chapters 1 and 2, Harter (1986, 1992) regards the false-self behaviors of adolescent girls as particularly problematic, inextricably linked, in a way distinct from boys, to girls' declining sense of self-worth throughout adolescence. Young girls more so than boys accept themselves contingent upon others accepting them, they develop a false-self. Bruch (1973) conceptualizes anorexia nervosa as the "desperate struggle for a self-respecting identity" (p. 250). These young women rely upon the ideas of others to explain their own behavior and are constantly "double-tracking ... pursuing one's own thought while trying simultaneously to figure out parental motives and reactions and continually monitoring others' responses" (p. 62). One of her clients says, "I really feel I am not myself and this is really sick, this not being myself, not my body or person, not really a human being" (p. 71). Bruch argued that the basic misperceptions of these women involve a fundamental alienation from self and others. As one woman states: "I have this unbelievable fear of people not liking me" together with its counterpart, "I'm not worthy o