Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Women's and men's networks in the workplace : attitudes, behaviours and outcomes McBain, Laura-Lynne 1990

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

Item Metadata

Download

Media
UBC_1991_A2 M32.pdf [ 16.14MB ]
[if-you-see-this-DO-NOT-CLICK]
Metadata
JSON: 1.0053629.json
JSON-LD: 1.0053629+ld.json
RDF/XML (Pretty): 1.0053629.xml
RDF/JSON: 1.0053629+rdf.json
Turtle: 1.0053629+rdf-turtle.txt
N-Triples: 1.0053629+rdf-ntriples.txt
Original Record: 1.0053629 +original-record.json
Full Text
1.0053629.txt
Citation
1.0053629.ris

Full Text

WOMEN S AND MEN S NETWORKS IN THE WORKPLACE: 1  1  ATTITUDES, BEHAVIOURS, AND OUTCOMES by LAURA-LYNNE McBAIN B.S.N., The University of British Columbia, 1975 M.A., The University of British Columbia, 1983  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF EDUCATION in THE FACULTY OF GRADUATE STUDIES (Department of Counselling Psychology)  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA November 1990 ® Laura-Lynne McBain, 1990  In presenting  this thesis in partial fulfilment 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  Counselling  Psychology  The University of British Columbia Vancouver, Canada  DE-6 (2/88)  ABSTRACT  Homosociality, the societal norm toward same-gender social bonding, has been hypothesized as an important explanatory variable in the maintenance of occupational segregation by gender and the low status of women in traditionally male-dominated occupations (LipmanBlumen, 1976; Reagan & Blaxall, 1976). In this investigation of homosociality in the workplace, 257 women and 197 men employed in managerial, supervisory, professional, and technical positions in seven organizations completed a questionnaire regarding their career development and interpersonal relationships in their current organization. Predictions derived from homosociality theory and the literature and research on mentoring, friendship, and organizational networks were tested. Of the 17 hypotheses associated with five research questions, 8 were fully or partially supported, 6 were not supported, and 3 could not be tested because factor analysis did not support the variable of interest (lifetime attachment). Alpha was apportioned using the Bonferroni inequality procedure; probability levels ranged from .025 to .0025 depending on the number of significance tests conducted for each question. Analysis of variance (Gender x Gender Composition of Network) and simple main effects analysis performed on mentoring and relationship provisions (intimacy, similarity, defiance of convention, respect for differences) scores indicated one significant main effect for gender: women's same-gender networks provided more intimacy than men's. Significant main effects for gender composition were: (a) men's same-gender networks provided more mentoring than their cross-gender networks; (b) women's same-gender networks provided more intimacy than their cross-gender networks; and (c) for both genders, same-gender networks provided higher levels of similarity and defiance of convention than cross-gender networks. Correlational analyses indicated: (a) for women, but generally not for men, homosocial attitudes were significantly related to the size and activities of same- and cross-gender networks; (b) for both genders, same- and cross-gender mentoring and primarily same-gender relationship provisions were positively and significantly related to career- and job-related outcomes.  i i i  Homosociality was evident in attitudes, network activities, and outcomes. Results also indicated signs of organizational gender integration. Implications for theory and counselling, and suggestions for future research, are discussed.  iv TABLE OF CONTENTS Abstract  ii  List of Tables  vii  List of Figures  x  Acknowledgement Chapter I. A.  B. C. Chapter II. A.  INTRODUCTION  1  Background to the Problem: Conceptual Foundations  1  1. 2. 3.  1 3 7 8 10  REVIEW OF THE LITERATURE  13  Homosociality  13  1. 2. 3.  13 16  E. F.  Theories of Male Bonding Theories of Female Bonding Research on the Behaviours and Special Characteristics of Same-Gender Relationships Homosocial Attitudes The Relationship Between Homosocial Attitudes and Behaviour  19 23 35  Cross-Gender Relationships Homosociality and Organizational Networks  38 40  1. 2. 3.  41 43  4. D.  Occupational Segregation by Gender Homosociality Homosociality and Organizational Networks  Statement of the Problem Significance of the Study  4. 5. B. C.  xi  Independent Achievement Versus Development in Relationship The Psychology of Tokenism Advancement Through the Activities of Informal Networks: Blocks and Opportunities Homosociality and Network Structure  46 50  Mentoring  59  1. 2. 3. 4.  60 70 71  Mentoring Functions Gender Differences in Mentoring Relationships Outcomes of Mentoring Behaviours A Network of Supportive Relationships as an Alternative to a Single Mentor  The Stages of Career Development Research Questions and Hypotheses  74 76 77  V  Chapter III. METHOD  90  A. Research Procedures  90  1. Recruitment of Participating Organizations 2. Generation of the Sample  90 91  3. Questionnaire Distribution and Return  92  B. Measures  95  1. The Close Social Relationships Scale (CSRS) 2. The Bonds Between Women/Bonds Between Men Scale (Bonds) 3. Supportive (Mentoring) Functions and Behaviours 4. Satisfaction With Progression 5. Average Annual Salary Increase 6. Average Number of Promotions per Year 7. Skill Development 8. Job Satisfaction : 9. Organizational Commitment 10. Acceptance by Co-workers 11. Loneliness at Work 12. Demographic Measures C. Design and Statistical Analysis Chapter IV. RESULTS  :  A. Characteristics of the Respondent Sample 1. Number of Respondents 2. Description of the Sample B. Hypothesis-Testing Analyses 1. Research Question 1: What is the relationship between an individual's homosocial attitudes and the size of his or her same- and cross-gender networks? 2. Research Question 2: What is the relationship between an individual's homosocial attitudes and the levels of supportive (mentoring) behaviours and relationship provisions he or she receives from same- and cross-gender networks? 3. Research Question 3: Are there gender differences in the levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks? 4. Research Question 4: How do homosocial attitudes in the interaction with levels of supportive (mentoring) behaviours received from same- and cross-gender networks relate to career- and job-related outcomes? 5. Research Question 5: How do levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks relate to career- and job-related outcomes?  96 101 106 111 111 112 112 114 114 115 116 117 117 125 125 125 127 128  129  133 143  152 161  vi  Chapter V. A.  DISCUSSION  170  Discussion of the Results for Each of the Five Research Questions  171  1. 2.  3. 4.  5.  B.  C.  D.  Research Question 1: What is the relationship between an individual's homosocial attitudes and the size of his or her same- and cross-gender networks? Research Question 2: What is the relationship between an individual's homosocial attitudes and the levels of supportive (mentoring) behaviours and relationship provisions he or she receives from same- and cross-gender networks? Research Question 3: Are there gender differences in the levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks? Research Question 4: How do homosocial attitudes in the interaction with levels of supportive (mentoring) behaviours received from same- and cross-gender networks relate to career- and job-related outcomes? Research Question 5: How do levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks relate to career- and job-related outcomes?  171  173 176  185 186  Limitations of the Study  189  1. 2. 3.  189 190 194  Individual Perceptions Characteristics of the Sample The Exploratory Nature of the Study  Implications and Recommendations for Future Research  197  1. 2.  197 204  Implications for Theory Implications for Counselling  Conclusions  . 208  REFERENCES  209  APPENDICES  220  Appendix A.  Measures  220  Appendix B.  Characteristics of the Respondent Sample  258  Appendix C.  Preliminary Analyses  265  vii LIST OF TABLES  Table 1: Table 2:  Table 3:  Table 4:  Table 5: Table 6: Table 7: Table 8: Table 9:  Table 10:  Table 11:  Table 12:  Table 13:  Bivariate Correlations Between Homosocial Attitudes and Size of Same- and Cross-Gender Supportive (Mentoring) Networks  131  Bivariate Correlations Between Homosocial Attitudes and Levels of Supportive (Mentoring) Behaviours and Relationship Provisions Received from Same- and Cross-Gender Networks (Women)  136  Bivariate Correlations Between Homosocial Attitudes and Levels of Supportive (Mentoring) Behaviours and Relationship Provisions Received from Same- and Cross-Gender Networks (Men)  137  Results of 2 x 2 ANOVAs Performed on Similarity, Defiance of Convention, and Respect of Differences Scores: Mean Scores and F-ratios  150  Results of 2 x 2 ANOVAs With Same-Gender IVs Performed on Careerand Job-Related Outcome Scores: Mean Scores and F-ratios (Women)....  155  Results of 2 x 2 ANOVAs With Same-Gender IVs Performed on Careerand Job-Related Outcome Scores: Mean Scores and F-ratios (Men)  156  Results of 2 x 2 ANOVAs With Cross-Gender IVs Performed on Careerand Job-Related Outcome Scores: Mean Scores and F-ratios (Women)....  158  Setup for 2 x 2 ANOVAs With Cross-Gender IVs: Cell Sizes and Mean Scores for Three Career- and Job-Related Outcome Variables (Men)  160  Standard Multiple Regression Results for Three Career- and Job-Related Outcomes Regressed Separately Onto Supportive (Mentoring) Behaviours and Relationship Provisions (Women)  163  Standard Multiple Regression Results for Three Career- and Job-Related Outcomes Regressed Separately Onto Supportive (Mentoring) Behaviours and Relationship Provisions (Men)  164  Bivariate Correlations Between Supportive (Mentoring) Behaviours and Relationship Provisions Received from Same- and Cross-Gender Networks and Three Career- and Job-Related Outcomes (Women)  165  Bivariate Correlations Between Supportive (Mentoring) Behaviours and Relationship Provisions Received From Same- and Cross-Gender Networks and Three Career- and Job-Related Outcomes (Men)  166  Network Members Considered While Completing the Supportive (Mentoring) Behaviours Measure: Numbers (and Percentages) of Same- and Cross-Gender Persons Listed Within Each Category of Relationship to Respondent  180  viii  Table B-1: Table B-2:  Number of Distributed, Returned, Excluded, and Retained Questionnaires  259  Age and Organizational Tenure of Respondents and Potential Respondents  260  Table B-3:  Levels of Education of Respondents  261  Table B-4:  Job Categories of Respondents  262  Table B-5:  Marital Status of Respondents  263  Table B-6:  Ethnic Designations of Respondents  264  Table C-1:  Descriptive Statistics for 10 Career- and Job-Related Outcome Measures (Women)  269  Descriptive Statistics for 10 Career- and Job-Related Outcome Measures (Men)  270  Interscale Correlation Matrices for the 10 Career-and Job-Related Outcome Measures  274  Oblique Primary-Factor Pattern Matrix for 10 Career- and Job-Related Outcome Variables (Women)  276  Oblique Primary-Factor Pattern Matrix for 10 Career- and Job-Related Outcome Variables (Men)  277  Oblique Primary-Factor Intercorrelations for the Items of the Bonds Between Women/Bonds Between Men Scale  281  Oblique Primary-Factor Pattern Coefficients for Bonds Between Women/Bonds Between Men Scale Items  282  Descriptive Statistics for the Individual Scales of the Bonds Between Women/Bonds Between Men Scale  286  Interscale Correlations for the Bonds Between Women/Bonds Between Men Scale  288  Table C-2: Table C-3: Table C-4: Table C-5: Table C-6: Table C-7: Table C-8: Table C-9: Table C-10: Table C-11: Table C-12: Table C-13:  Oblique Primary-Factor Pattern Matrix and Communalities (h*-) for the Five Individual Scales of the Bonds Between Women Scale  289  Oblique Principal Component Pattern Matrix for the Three Individual Scales of the Bonds Between Men Scale  290  Oblique Primary-Factor Intercorrelations for 30 Items of the Close Social Relationships Scale (CSRS)  298  Oblique Primary-Factor Pattern Coefficients for 30 Items of the Close Social Relationships Scale (CSRS)  299  ix Table C-14: Table C-15: Table C-16: Table C-17:  Descriptive Statistics for the Individual Scales of the 30-ltem Close Social Relationships Scale (Same-Gender Relationships)  302  Descriptive Statistics for the Individual Scales of the 30-ltem Close Social Relationships Scale (Cross-Gender Relationships)  303  Interscale Correlations for the 30-ltem Close Social Relationships Scale (CSRS)  304  Descriptive Statistics for the Supportive (Mentoring) Behaviours Measure  306  X  LIST OF FIGURES Figure 1:  Overall model of variables and research questions  Figure 2:  Variables, measures, design, and analysis for Research Question 1  118  Variables, measures, design, and analysis for Research Question 2  119  Variables, measures, design, and analysis for Research Question 3  120  Variables, measures, design, and analysis for Research Question 4  121  Variables, measures, design, and analysis for Research Question 5  122  Graphic presentation of 2 x 2 between-within A N O V A with one between subjects factor (gender of respondent) and one within subjects factor (gender composition of network) performed on Total Supportive Behaviour Scores  144  Graphic presentation of 2 x 2 between-within A N O V A with one between subjects factor (gender of respondent) and one within subjects factor (gender composition of network) performed on Intimacy scores  148  Figure 3:  Figure 4:  Figure 5:  Figure 6:  Figure 7:  Figure 8:  11  xi ACKNOWLEDGEMENT  I wish to thank the members of my Supervisory Committee - Richard Young (Chair), Ralph Hakstian, Bonnie Long, Dan Perlman, Craig Pinder, and Lorette Woolsey -- for sharing their knowledge and experience with me. I also would like to express my gratitude to the contact people and individual respondents in the seven organizations from which the data for this study were collected. Funding sources for my doctoral programme and for my dissertation research are noted on the biographical form at the end of this document. In addition, I acknowledge my deep appreciation of the many colleagues, friends, and family members too numerous to mention by name, for being consistent sources of encouragement and practical help. Special thanks to Klaus Schroeder for generously sharing his opinions, experience, and measuring instruments; to Marsha Schroeder, Jessica McFarlane, and Tim Harpur for demystifying statistical analysis; to Jeeva Jonahs for her cheerful and careful approach to the word processing of this document; and to Maureen Ashfield for endless hours of proofreading and emotional support. Finally, I honour and thank the "wild patience" (Rich, 1981, p. 8) that sustained me during the years of work on this project.  1 CHAPTER I. INTRODUCTION  The purpose of this study was to test theories of male and female homosociality (samegender social bonding) in the organizational context. The character of men's and women's relationships was examined by measuring the mentoring behaviours and relationship provisions supplied by same- and cross-gender relationship networks in seven different organizations. In addition, the relationship between homosocial attitudes and network activities was studied. Finally, measures of career- and job-related outcomes such as promotions, skill development, and job satisfaction were utilized in the investigation of the relationships between homosociality and these outcomes. The study is introduced in this chapter with an overview of the conceptual foundations which led to the problem statement, an expanded statement of purpose and listing of the specific questions that were addressed in this study, and a discussion of the significance of the study.  A. BACKGROUND TO THE PROBLEM: CONCEPTUAL FOUNDATIONS  1. Occupational Segregation by Gender The division of labour by gender is firmly entrenched in history as evidenced, for example, by the Judeo-Christian story of creation. After Adam and Eve had eaten the forbidden fruit from the tree of the knowledge of good and evil, their Creator cursed them and assigned them different roles: To the woman he said: "I will increase your labour and your groaning, and in labour you shall bear children. You shall be eager for your husband, and he shall be your master." And to the man he said: "Because you have listened to your wife and have eaten from the tree which I forbade you, accursed shall be the ground on your account. With labour shall you win your food from it all the days of your life You shall gain your bread by the sweat of your brow  2 until you return to the ground; for from it you were taken." (Genesis 3:16,17,19a, The New English Bible, 1970, p. 4)  The enactment of these gender roles has continued into modern times; in 1990, many women and men continue to operate in largely separate spheres of activity. For example, since the turn of the twentieth century, occupational segregation by gender has persisted in Canada (Moore, 1985) and the United States (Reskin & Hartmann, 1986) despite markedly increased participation rates of women in higher education and the paid labour force (Blau, 1984; Canada, Women's Bureau, 1987; Statisics Canada, 1989), a significant rise in public and governmental awareness of the status and rights of women (Ornstein, 1982), and strengthening employment equity legislation (Beller, 1984; Employment Equity Act, 1986). Occupational segregation by gender has two facets. First, in horizontal segregation (Hakim, 1979; Moore, 1985), women and men are concentrated in different occupations. Women tend to work in a small number of occupations in which the majority of workers are female, while men tend to work in a larger number of occupations in which the workers are predominantly male (Reskin & Hartmann, 1986). Second, in vertical segregation (Hakim, 1979; Moore, 1985), women tend to be located in occupations significantly lower in prestige, authority, or income than are men. Many explanations have been suggested for occupational segregation by gender. Theories of labour market discrimination summarized by Blau (1984) focus on discrimination against women by employers, co-workers, or customers. The human capital model, also summarized by Blau (1984), emphasizes that the traditional division of labour by gender within the family has resulted in women voluntarily choosing to enter occupations requiring less commitment and training because they expect a shorter and more discontinuous involvement in the paid labour force than men. Lipman-Blumen (1976) offers still another explanation. In her homosocial theory of sex roles, she presents a conceptual framework for understanding the gender segregation which pervades all of our social institutions - the workplace, the home and family, the legal system, the political arena, and the sphere of recreation. Briefly, homosociality is defined as "the seeking, enjoyment, and/or preference for the company of the same sex" (Lipman-Blumen, 1976, p. 16). This prevalent social  3 norm toward same-gender social bonding has been hypothesized to be an explanatory variable in occupational gender segregation and in the low status of women in male-dominated professions (Lipman-Blumen, 1976; Reagan & Blaxall, 1976). While theories of labour market discrimination and the human capital model have been tested to some degree, to date homosociality theory has not been tested empirically in the organizational context. Blau (1984), upon summarizing the empirical investigations of labour market discrimination theory and the human capital model, concluded that: It appears that sex segregation in employment is an important mechanism for producing sex differences in earnings and that the occupational differences between men and women do not seem to be consistent with optimizing behavior on the part of women. However, considerably more work is needed to understand the causes of sex differences in occupational distributions fully and to determine the role of such occupational differences in producing male-female pay differentials, (p. 139) Homosociality was posited as one heretofore uninvestigated factor in gender differences in occupational distributions. The current study addressed this deficit by examining patterns of homosociality in women's and men's relationships in the workplace. In the next section, homosociality is defined further.  2. Homosociality  Because homosociality is a relatively new field of study, it is fraught with theoretical gaps and inconsistencies. In this section, divergent definitions are synthesized to provide a conceptual framework for the study of homosociality within the organizational context. A more complete review of homosociality theory and research follows in Chapter II (Review of the Literature). Homosociality has two distinctly different dimensions. First, homosociality may refer to actual relationships between same-gender persons, to "the social ties or bonds that women have with other women and that men have with other men" (Woolsey, 1987a, p. 117). These actual samegender relationships may be different in character for men and for women, as suggested by  4 Bernard's (1976) definition of homosociality as "the different ways that men and women relate to their own sex" (pp. 227-228). In the current study, the activities and character of actual relationships were examined by measuring the mentoring behaviours and relationship provisions supplied to individual respondents by same- and cross-gender networks in their particular workplace. While the provisions of crossgender networks obviously are not aspects of same-gender bonding, information about how crossgender relationships differ from same-gender relationships will help to define the construct of homosociality. Definitions and units of measurement for mentoring behaviours and relationship provisions are introduced briefly below. Mentoring has been defined as a "relationship in which a person of greater rank or expertise teaches, counsels, guides and develops a novice in an organization or a profession" (Alleman, 1986, p. 45). A mentoring relationship may provide any of the following mentoring functions to a protege: Sponsoring, Exposure and Visibility, Teaching the Job, Teaching the Informal System, Protection, Role Modeling, Encouragement, and Personal Counselling (Schroeder, 1988/1989). Specific mentoring behaviours associated with each of these functions may be provided by persons other than a mentor in the traditional sense (Kram, 1985; Schroeder, 1988/1989). The present study examined the mentoring behaviours provided by a network of support persons within each respondent's organization. The total level of mentoring behaviours supplied by each of two networks (same-gender and cross-gender) were the units of measurement in the present study. The words supportive and mentoring both are used in the current document to describe the behaviours associated with mentoring functions provided to respondents by same- and crossgender networks. This terminology was adapted from that of Schroeder (1988/1989) who, for the purpose of his study, renamed mentoring behaviours as supportive behaviours and referred to the overall level of mentoring behaviours measured by his scale as the Total Supportive Behaviour Score. Schroeder did this to acknowledge two things. First, what some writers have referred to as mentoring functions are conceptually comparable to what other authors have called work-related social support. Second, as discussed above, mentoring functions may be provided by co-workers  5  and others who are not necessarily seen as mentors. As Schroeder's scale was the measure of mentoring behaviours utilized in the current study, and because the level of mentoring behaviours provided by relationship networks that may or may not have included traditional mentors was measured, the current writer used the term supportive (mentoring) behaviours when this was necessary for clarity. When using the term supportive was not necessary for clarity, only the term mentoring was used. Relationship provisions were defined as the core aspects of close social relationships. Relationships that provide mentoring behaviours also can supply relationship provisions but they do not necessarily do so. Mentoring relationships that provide only career functions (e.g., Sponsoring, Exposure and Visibility, Teaching the Job, Teaching the Informal System, Protection) are more taskoriented and less emotionally close than are relationships that also provide psychosocial functions (e.g., Role Modeling, Encouragement, and Personal Counselling) (Kram, 1985). The relationship provisions measured in the current study were: Intimacy, Similarity, Defiance of Convention, Emotional Support, Respect for Differences, and Lifetime Attachment (Woolsey, Hakstian, & McBain, 1988b). While certain aspects of the psychosocial provisions of mentoring relationships were included in the overall measure of mentoring behaviours used in the present study (Schroeder, 1988/1989), the additional measurement of relationship provisions provided a more complete assessment of the psychosocial aspect of these relationships. Homosociality does not refer only to the activities and character of actual same-gender relationships. The second dimension of homosociality involves an "individual's affect, behaviour and cognitions regarding the same sex in general" (Woolsey, 1987a, p. 117). In the current study, this second aspect of homosociality was referred to as homosocial attitudes. The results of factor-analytic work on the Bonds Between Women/Bonds Between Men Scale (Woolsey, Hakstian, & McBain, 1988a), the measure of homosocial attitudes adapted for use in the current study, indicated that homosocial attitudes had a five-factor structure for women and a three-factor structure for men. Conceptual analysis of the factors suggested that the three factors common to women and men -- Cross-Gender Preference, Same-Gender Preference, and Valuing --  6  could be broadly subsumed under Lipman-Blumen's (1976) definition of homosociality as "the seeking, enjoyment, and/or preference for the company of the same sex" (p. 16). The item content of the Cross-Gender Preference factor suggested that a person with a high level of Cross-Gender Preference would prefer that persons of higher status or authority be of the other gender. The item content of the Same-Gender Preference factor indicated that someone with a high level of SameGender Preference would prefer that his or her peers (e.g., co-workers, teammates) be of his or her own gender. The Valuing items suggested that same-gender relationships can be valued in and of themselves, because they contribute something important to one's life. The final two homosocial attitude factors - Identification and Jealousy -- were supported only by the results for women; factor analysis of the same items using data gathered from men did not indicate the presence of Identification or Jealousy factors for men (Woolsey et al., 1988a). Rawlings and Carter's (1977) adaptation of Allport's (1955) minority group theory was utilized by Woolsey et al. (1988a) in their conceptual analysis of the women's Identification and Jealousy factors. The item content of these two factors suggested that they could be conceptualized as aspects of the ego defenses developed by women as a group to counteract the effects of prejudice and discrimination. These ego defenses are similar to those found in other disadvantaged social groups. Following Rawlings and Carter (1977), Identification was classified as an "extropunitive" ego defense; that is, the anger that results from unjust treatment is directed outward. Extropunitive ego defenses tend to strengthen ties between women (Rawlings & Carter, 1977). The item content of the Identification factor suggested that a woman with a high level of Identification would indicate a clear awareness of the cultural prejudice and discrimination experienced by herself and other women and would indicate an intent to assist other women in overcoming these disadvantages. If this high level of Identification were to be translated into behaviour, one might see this woman making a special effort to help other women get ahead in the workplace. The homosocial attitude of Jealousy was classified as an "intropunitive" ego defense. The hostility that results from unjust treatment is directed inward, toward the self and others like the self (i.e., other women), causing feelings of jealousy, competition, and threat between women, all of  7 which tend to erode ties with other women (Rawlings & Carter, 1977). If translated into behaviour, a woman with a high level of Jealousy toward other women might experience difficulties in samegender relationships, neither giving to nor receiving from other women the supportive mentoring behaviours which have been linked to career development. The proposed study was an investigation of both of the previously noted aspects of homosociality - the behaviours and character of actual relationships and homosocial attitudes.  In the next section, the hypothesized effects of homosociality on organizational networks are reviewed briefly.  3. Homosociality and Organizational Networks  The exclusion of women from the informal networks of men in the workplace has been cited repeatedly as a key factor in women's slow entry into positions of power within organizations (Abramson, 1979; British Columbia, Ministry of Industry and Small Business Development [MISBD], 1979; Forisha & Goldman, 1981; Hennig & Jardim, 1977; Kanter, 1977; Lincoln & Miller, 1979; Lovelady-Dawson, 1981; Riger & Galligan, 1980; Viega, 1976). One result of women's exclusion from men's informal networks has been a deficiency in career sponsorship and mentoring for women (Bolton, 1980; Collins, 1983; Epstein, 1970; LaFrance, 1982; Riger & Galligan, 1980). While "women's networks" have become increasingly common over the past decade as an alternative method of career support for women (Forisha & Goldman, 1981; Fraser & Hendry, 1982), women's same-gender relationships in male-dominated organizations may present special problems of their own. For example, some upwardly mobile women in male-dominated systems have tended to dissociate themselves from other women, identifying and associating with men instead (Dellacava & Engel, 1979; Laws, 1975; Marshall, 1984). Eichenbaum and Orbach (1988), both psychotherapists, wrote a book about women's friendships based on the personal and clinical insights they had gained in working with women and in women's groups since the 1970s. They suggest that issues of jealousy, competition, and envy between women seem to be on the increase along with the rapidly changing gender roles of the last decade. Woolsey and McBain's (1987a) clinical observations and  8 analyses of conflicts experienced in all-woman work groups led them to conclude that when power issues and feelings of jealousy between women are addressed openly, healthy growth and shared empowerment may ensue. If allowed to fester, though, power struggles between women may result in serious reductions in the quality and effectiveness of women's work relationships. Women in today's organizations, then, would appear to be dealing with two issues related to network activities - the difficulty of becoming integrated into men's networks and the potential for problems in maintaining supportive relationships with other women. These issues may have important implications for the advancement process. The findings of a British Columbia, Ministry of Industry and Small Business Development (1979) study suggested that the skills of advancing one's work are learned within the informal network of relationships embedded in the structure of organizations. It would seem reasonable to conclude, then, that difficulties in work relationships may impede career development. One of the conclusions of the British Columbia, Ministry of Industry and Small Business Development (1979) study was that the social character of the organizational process must be addressed as a legitimate area of organizational research in order to move toward truly equal access to career advancement for all persons. In the next section, the problem and research questions that were investigated in the current study are outlined.  B. STATEMENT OF THE PROBLEM As indicated by the literature reviewed earlier in this chapter, traditional patterns of occupational segregation by gender may preclude employment equity for women. Aspects of the prevalent tendency toward homosociality may contribute to the maintenance of vertical segregation by gender within organizations. To date, there have been few studies of the men's and women's networks in organizations. Brass (1985) and Lincoln and Miller (1979) examined the structural characteristics of relationship networks in organizations but did not investigate the activities within the networks. Studies of mentoring (e.g., Alleman, 1986; Alleman, Cochran, Doverspike, & Newman, 1984; Kram, 1985; Schroeder, 1988/1989) have addressed the specific activities of mentoring but  9 have not explored in detail the potentially different relationship styles and needs of women and men. To date, only a limited number of empirical investigations have addressed the issue of gender differences in mentoring among managers (Ragins, 1989). Research on friendship has identified distinct patterns of difference between men's and women's friendships (Aries & Johnson, 1983; Bell, 1981; Caldwell & Peplau, 1982; Sherrod, 1989; Woolsey & McBain, 1987b). However, these investigations have focused on friendships outside the workplace. There are no known studies that specifically have addressed the relationship between homosocial attitudes and network activities. The primary purpose of this study was to investigate patterns and potential outcomes of homosociality within organizations by: 1.  Investigating the relationship between homosocial attitudes and network activities.  2.  Examining gender differences in the mentoring behaviours and relationship provisions supplied by same- and cross-gender networks (i.e., investigating the actual relationships aspect of homosociality).  3.  Examining the relationship between homosocial attitudes and behaviours and indices of employee outcomes (career- and job-related variables such as promotions, skill development, job satisfaction).  The networks were examined from an ego-centered perspective (Alba, 1982). That is, the network data that were collected pertained to small, distinct relationship networks centred around individual respondents in the sample. Respondents were requested to provide various types of information about their own networks. The network structure of the larger organization was not studied. Five general research questions were posed to address key aspects of the problem outlined above: 1. What is the relationship between an individual's homosocial attitudes and the size of his or her same- and cross-gender networks?  10 2.  What is the relationship between an individual's homosocial attitudes and the levels of supportive (mentoring) behaviours and relationship provisions he or she receives from same- and cross-gender networks?  3.  Are there gender differences in the levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks?  4.  How do homosocial attitudes in interaction with levels of supportive (mentoring) behaviours received from same- and cross-gender networks relate to career- and jobrelated outcomes?  5.  How do the levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks relate to career- and job-related outcomes?  The hypotheses which correspond to each of these research questions are presented at the end of Chapter II (Literature Review). The overall model of variables and research questions is presented in Figure 1.  C. SIGNIFICANCE OF THE STUDY The Employment Equity Act (1986) requires all federally regulated employers of 100 or more persons to carry out Employment Equity strategies and to provide annual reports on their results. The purpose of the Act is: to achieve equality in the work place so that no person shall be denied employment opportunities or benefits for reasons unrelated to ability and, in the fulfilment of that goal, to correct the conditions of disadvantage in employment experienced by women, aboriginal peoples, persons with disabilities and persons who are, because of their race or colour, in a visible minority in Canada. (Employment Equity Act, 1986, p. 1)  The Act further states that employment equity "means more than treating persons in the same way but also requires special measures and the accommodation of differences" (p. 1). The results of this investigation of homosociality - of (a) the tendency of people to prefer and to seek social relationships with same-gender persons, and (b) the different styles of same-  Homosocial attitudes  2  Size of same- and cross-gender networks  Supportive (mentoring) behaviours and relationship provisions received from networks  Career- and Job-related outcomes  3  Gender  Figure 1. Overall model of variables and research questions.  Gender composition of networks (same-gender, cross-gender)  (Five research questions were posed to address different aspects of the model. Question  numbers corresponding to the particular research questions being addressed in each portion of the model have been placed on the appropriate connecting lines between the variables. In the case of Research Question 4, only the interaction between homosocial attitudes and supportive [mentoring] behaviours was Investigated.)  12 gender relationships for women and men -- provide valuable prerequisite information for designing programs to accommodate the differences between women and men in the workplace. This information could be used to "correct the conditions of disadvantage in employment experienced by women" (Employment Equity Act, 1986, p. 1), conditions which may result from their exclusion from the networks in which skills and support for career advancement seem to be learned and attained. Much of the tendency toward occupational segregation by gender may be caused by variables operating at an unconscious level in individual men and women, at a level untouched by legislation such as the Employment Equity Act. Homosociality may be one such variable. For example, the preference for males as work associates by both men and women (Woolsey-Toews, 1975) may be related to stereotyped images of gender roles and to unconsciously held beliefs about what is desirable in terms of work relationships with same- and cross-gender persons. At this point in history, it would seem to be important to study homosociality in order to determine its key aspects in the work relationships of women and men. Otherwise, "even the well-intentioned person [in terms of employment equity] is likely to behave in ways that negate those intentions without any awareness of having done so" (Woolsey-Toews, 1975, p. 31). Noe (1988), upon reviewing the research related to women and mentoring, concluded that "few rigorous quantitative studies of the antecedents or outcomes of mentoring have appeared in the organizational behavior literature" (p. 66). The present study addressed this deficit by examining a set of variables (homosocial attitudes) conceptualized as antecedents of mentoring and several variables (career- and job-related variables such as promotions, skill development, job satisfaction) conceptualized as outcomes of mentoring.  13 CHAPTER II. REVIEW OF THE LITERATURE  In this chapter, the theory and research from which the present study emerged are reviewed. The areas reviewed are: homosociality, cross-gender relationships, homosociality and organizational networks, mentoring, and the stages of career development. The hypotheses which emerged from this literature review are presented at the end of the chapter.  A. HOMOSOCIALITY  Theories of male and female homosociality most directly related to the organizational context are reviewed in the next two subsections. Following that, research on actual same-gender relationships is reviewed.  1. Theories of Male Bonding  Four theories are reviewed in this subsection. The first two theories reviewed here -- those of Lionel Tiger (1969) and Jean Lipman-Blumen (1976) -- are very similar. Both suggest that males bond in hierarchical groupings with issues of power, control, and status being the main facets around which bonding occurs. Tiger's (1969) theory is that males have a biologically based predisposition to form strong group bonds with one another "in a variety of situations involving power, force, crucial or dangerous work, and relations with their gods  [and to] consciously and emotionally exclude females from  these bonds" (p. 112). According to Tiger, current patterns of male bonding reflect and have arisen from the historical male roles of hunter and warrior, roles which necessitated cooperative bonding among men for the purpose of successful hunting and defense. From these early patterns of male bonding evolved the social institutions which formed the backbone of human communities. Tiger's examination of anthropological data on many social institutions - armies, sports organizations, religious structures, politics, secret societies, family systems, and economics -- led him to the conclusion that the public forum is a male forum dominated by hierarchies of adult males. He  14 described the bonds between males as "the spinal column of a community, in this sense: from a hierarchical linkage of significant males, communities derive their intra-dependence, their structure, their social coherence, and in good part their continuity through the past to the future" (p. 60). Tiger (1969) noted that, in contrast to males, females rarely dominate authority structures and seem to lack the strong same-gender bonding tendency evident in males. On the basis of numerous cross-cultural studies of female political behaviour, he concluded that female organizations are relatively obscure and unimportant to the "political structure of their communities and the establishment of a dominance hierarchy (p. 72). While he noted that women do form groups "for some purposes, such as childcare, gathering, and farming, or for simple gregariousness" (p. 72), his research lead him to the conclusion that women's same-gender bonds do not persist over time as do men's. Lipman-Blumen (1976) also notes the widespread presence of male dominance hierarchies from which women are excluded. Unlike Tiger, though, Lipman-Blumen suggests a sociological rather than a biological basis for male bonding and for the exclusion of women from these bonds. Lipman-Blumen speculates that the roots of male bonding are in the historical male roles of hunter and warrior, roles closed to women because of their primary childbearing and child-nursing activities. Hunter and warrior roles placed men in charge of acquiring and protecting the fundamental resources of their society -- territory and food. The dominance hierarchies which developed from these roles have continued into technological times which no longer require such a strict differentiation and stratification of male and female roles. Lipman-Blumen (1976) contends that, beginning in childhood, males are interested in, attracted to, and stimulated by other males. This process is encouraged by the social stratification process which classifies individuals and groups according to their perceived value to society. Men, perceived as being more valuable to society than are women, are systematically placed in more prestigious roles than women. The result of this stratification process is that, until recently, men have been placed in positions of exclusive access to and control over economic, political, educational, legal, and social resources, while women have been restricted to a more limited range  15 of resources -- service, youth, beauty, sexuality, and parenthood. Men, because of their greater resource control, have been more attractive and useful to other men, and to women as well. Men identified with and sought assistance from other men; women, recognizing their own paucity of resources in comparison to men, also looked to men, rather than to other women, for protection and help. Lipman-Blumen (1976) suggests that the homosocial world has existed primarily for the benefit of men; the powerful "old-boy network" of the male world has been an important aid to the control of resources. Women, with relatively little to bargain with in resource exchange relationships that exist beyond the domestic sphere, have had to derive their status indirectly through their relationships with men. While Lipman-Blumen notes the recent development of women's networks outside the traditional domestic realm, she suggests that such networks are still in their infancy. Pleck (1975) takes the theories of Tiger (1969) and Lipman-Blumen (1976) one step farther. He notes the historical pattern of male bonding in which all-male institutions and activities supported male solidarity while devaluing and excluding women. He then focuses on the impact of gender role socialization on men's same- and cross-gender relationships. Summarizing research on male homosocial relationships, he states that "male sociability is closely connected with male sex role training and performance and is not characteristically a medium for self-exploration, personal growth, or the development of intimacy" (p. 235). The social needs met by men's relationships with men, then, "must be met in ways that support male role performance and not in ways that might threaten it" (p. 234). For example, the male executive must have relationships with male clients and co-workers which enable him to "transact business with them and get the emotional support he needs himself, but these relationships cannot be so intimate as to inhibit the competition or selfaggrandizement required of him" (p. 234). Pleck (1975) suggests that the gender role revolution (referred to as the sex role revolution in 1975) is ushering in the possibility of a new type of brotherhood between men, a brotherhood that "transcends both older styles of male solidarity, usually limited in its own right and based on the devaluation of women, as well as more recent male styles of channeling intimacy needs exclusively  16 into relationships with women" (p. 242). He suggests that men, as well as women, have suffered the consequences of limited gender role prescriptions and that they, like women, can draw understanding and strength from others of their own gender as they attempt to come to grips with common issues and experiences as men. This type of male bonding is based on a sense of commonality and support for other men, comparable to that found between feminist women. Pleck (1975) quotes the words of one man discovering this type of bonding with other men. After attending a party given by a group of feminist women, he said: The women there were really high together. The way they were dancing and interacting with each other, it was like they were really celebrating each other, in a beautiful way. And then I realized that thaf s what I wanted with other men, that same sense of celebration and affirmation of ourselves. I never thought that was possible for men, but I know it is now. (p. 241) Sherrod (1987) echoes many of Fleck's (1975) words in his discussion of the economic and social forces which have shaped the nature of men's relationships. Sherrod suggests that the economic forces of the industrial revolution precipitated two major changes in men's friendships. The first was a movement from the norm of shared reliance on kinship-friendship networks in rural villages to the norm of direct competition with other men for work and wages. The second was a concomitant shift from reliance on larger social networks for support to a reliance on women (i.e., wives) for emotional support. Current economic and social changes that are altering women's lives and the structure of the family (e.g., continually increasing participation rates of women in employment outside the home, dual careers, single life styles) may indirectly affect men by forcing them to form different types of bonds with one another. Sherrod (1987) predicts a "resurgence of friendship and intimacy among men" (p. 238).  2. Theories of Female Bonding Theorists of female bonding state that the basis for women's same-gender bonding is affiliation and attachment, a pull toward mutuality in expressive and intimate relationships (Bernard, 1976,1981; Miller, 1976; Smith-Rosenberg, 1975).  17 An historical perspective on female bonding is provided by Jessie Bernard (1981), who wrote extensively about the female homosocial world of the nineteenth century. This world was characterized as the realm of the home and "the heart", a realm in which women, socialized to be specialists in the area of emotions and domestic relationships, took charge while men operated in the separate and larger male sphere of politics and economics. The female world of the nineteenth century offered invaluable support and validation for women. In a society in which women were barred from the dominant male sphere, women had a valued and secure place in their own sphere; women had power and a sense of belonging in their separate world. They were able to meet the majority of their relational needs in intimate and long-lasting relationships with other women, with whom they spent most of their time (Smith-Rosenberg, 1975) - needs which were not met by men because of their socialization toward inexpressivity and coolness. While this traditional women's sphere of "the heart" has eroded in the twentieth century because of geographic mobility, changing gender roles, and an emphasis on heterosexual relationships for both women and men (Bernard, 1981; Pleck, 1975; Smith-Rosenberg, 1975), present-day researchers continue to find evidence of more emotional richness and intimacy in women's same-gender relationships and a greater emphasis on collegiality and task-orientation in men's (Aries & Johnson, 1983; Bell, 1981; Booth, 1972; Caldwell & Peplau, 1982; Woolsey & McBain, 1987b). Writers such as Tiger (1969) and Lipman-Blumen (1976) failed to tap into the complex and emotionally rich world of relationships between women, presumably because their theories were written from the perspective of traditional male bonding styles, and women's relationships were examined and described reactively. For example, the questions asked at a 1974 symposium (Tiger & Fowler, 1978) that was organized to focus on female social systems, were so androcentric that they nearly obscured direct observations of female behaviour. In the anthropological/sociological studies reported at the symposium, the investigators had studied women to see if they fit a male norm of behaviour, the norm of social stratification according to dominance. The conclusions were that all-female hierarchies, or all-female organizations that are largely female managed, are difficult to find in modern society; that the tendency to form hierarchies may be a trait more common among  18 males than among females; and that when females do organize themselves according to hierarchical structures, that these forms may be different from those of men. While the information about female social behaviour per se that was gained at this symposium was minimal, Tiger (1978) did conclude that further study of the cooperative social structures evident among females was warranted. Writers and researchers such as Jean Baker Miller (1976) and Maccoby and Jacklin (1974) have done just that, noting that women as a group have more highly developed abilities than men in the areas of affiliation, cooperation, and dealing with the emotional side of life. This is not, Miller says, because women are more saintly than men, but because life (or gender role socialization) has led women to the position of being more practiced in cooperation and more able "to seek out and enjoy situations that require that quality" (p. 43). Miller (1976) states that a central feature of women's psychology is that women "stay with, build on, and develop in a context of attachment and affiliation with others" (p. 83). While this quality of "dependency" on relationships has been labelled problematic by theorists writing from a traditionally male perspective which values autonomy, competition, and hierarchical relationships based on status (Gilligan, 1982; Miller, 1976), Miller has framed women's affiliative focus as a strength. She and others (Eichenbaum & Orbach, 1983b) suggest that all persons -- both women and men - require significant attachments to others in order to develop. The growing body of literature and research on mentoring and peer relationships in organizations (e.g., Alleman, 1986; Alleman, Cochran, Doverspike, & Newman, 1984; Burke, 1984a, 1984b; Clawson, 1980; Collins & Scott, 1978; Klopf & Harrison, 1981; Kram, 1985; Kram & Isabella, 1985; Ragins, 1989) supports this suggestion. Research on relationship networks in the workplace (British Columbia, MISBD, 1979) indicates that "people do not advance by themselves. They are advanced" (p. 20) by other persons in the workplace with whom they have significant relationships. Men and women learn the skills necessary for career advancement in the context of relationships with superiors and others "who have pushed and challenged them and taught them to handle the organization" (British Columbia, MISBD, 1979, p. 20).  19 Miller (1976) contends that "at the present time, men are not as prepared to know this" (p. 83), to acknowledge that relationships are the means by which individual development proceeds. In contrast, a woman's "sense of self becomes very much organized around being able to make and then to maintain affiliations and relationships" (p. 83). Traditionally, then, women have been the specialists in the realm of relationships; they have lived out their lives primarily in the sphere of the home and "the heart" (Bernard, 1981). While women's increasing entry into the paid labour force has changed the context in which many women relate to each other and to men, Woolsey and McBain (1987a) argue that women still tend to bring a "depth of feeling and self-disclosure to their relationships wherever they occur, the work setting included" (p. 585). Present-day working women, and the organizations in which they work, are presented with a challenge: to create and maintain work relationships that are flexible enough to allow women to utilize their affiliative strengths as well as to exercise increasingly the achievement and task-oriented strengths (traditionally more characteristic of men) that are necessary for significant career advancement.  3. Research on the Behaviours and Special Characteristics of Same-Gender Relationships Most of the research related to homosociality at the level of actual relationships presented in this section has been drawn from the research on same-gender friendships. The work of Carroll Smith-Rosenberg (1975) is a well-documented source of information about the richness, complexity, and activities of women's relationships in the nineteenth century. Her work provides some background to twentieth century changes in patterns of same- and crossgender relationships. Smith-Rosenberg examined the diaries and correspondence of women and men in 35 middle class American families. Her extensive data base included thousands of letters written to women friends, family, spouses, and children between the 1760s and the 1880s. Some family collections were written over entire life spans; one collection included over 100,000 letters in addition to account books and diaries. Smith-Rosenberg (1975) applied information about the social context of mid-eighteenth and mid-nineteenth century America to the analysis of these writings, suggesting that it was this larger  20  milieu that supported the development of close emotional relationships between women. American society as a whole, and American families in particular, were characterized by strict gender role differentiation, which resulted in distinct female and male spheres. These different worlds were "inhabited by human beings with different values, expectations, and personalities" (p. 9). Contact between men and women was limited and formal, while contact between women was frequent and natural, because of the shared experiences of women's lives: The roles of daughter and mother shaded imperceptibly and ineluctably into each other, while the biological realities of frequent pregnancies, childbirth, nursing, and menopause bound women together in physical and emotional intimacy. It was within just such a social framework, I would argue, that a specifically female world did indeed develop, a world built around a generic and unself-conscious pattern of single-sex or homosocial networks. These supportive networks were institutionalized in social conventions or rituals which accompanied virtually every important event in a woman's life from birth to death, (p. 9) A woman's sphere was inhabited by other women and by children; it was bounded by home, church, and social visits to the homes of other women. Days and weeks could be spent almost entirely with other women, as women helped each other with domestic tasks, shopped together, vacationed together, talked, or shared times of sickness, trouble, or sorrow. Friendships formed within highly integrated networks held together by an inner core of kin -mothers, daughters, sisters, aunts, cousins and nieces. Women played a central role in each others' emotional lives; women felt contented and joyful when they were together and experienced isolation and despair when they were separated geographically. Within this secure and understanding world of women, "women could share sorrows, anxieties, and joys, confident that other women had experienced similar emotions" (p. 14). They were valued by each other; women who had little status or power in the male sphere had status and power with other women. Research that considers present-day relationships presents a similar picture. Booth (1972) compared men's and women's participation in friendships, kin relationships, and voluntary organizations in a test of Tiger "s (1969) theory that male same-gender bonds are stronger and more stable than those of women. The data he gathered from 800 American adults 45 years of age and older did not lend any support to Tiger's theory. In fact, the results of Booth's study led him to  21 conclude that it was women who seemed to "have the greater propensity for same-sex bonding" (p. 186). Overall, there were no gender differences in the number of close friends named by the women and men in Booth's study. Both women and men reported that the majority of their close friends were people of the same gender. However, men reported more cross-gender friends than did women. Women's friendships were affectively richer than those of men, as measured by spontaneity of contact and exchange of confidences. As stated by Booth (1972), "Women did things on the spur of the moment with 59 percent of their friends, men with 45 percent.... Women confided in 52 percent of their friends, men in 38 percent" (p. 186). Contrary to Tiger's (1969) theory, Booth found that women were as likely as men to belong to all-female organizations and to show stable patterns of participation in voluntary associations throughout their lives. While men joined more voluntary associations, women committed more time to these endeavours. Men tended to participate in groups of a more instrumental nature (e.g., voluntary organizations associated with economic, political, and military pursuits), while women tended to participate in groups more expressive in nature (e.g., educational, religious, recreational, health- and welfare-related groups). Women led the expressive groups more often than did men, while men assumed leadership of the instrumental groups more often than did women. Finally, in terms of kinship ties, Booth found that women maintained more family ties and had stronger bonds with family than did men. He concluded by noting that "these sex differences probably will diminish in the years to come as sex-role distinctions diminish. The advance in women's civil rights is signaling the change" (p. 192). Despite Booth's (1972) prediction of diminishing gender differences in relationship style, researchers continue to find evidence of more affective richness and intimacy in women's samegender relationships and a greater emphasis on activities and shared interests in men's (Bell, 1981; Caldwell & Peplau, 1982; Reis, Senchak, & Solomon, 1985; Woolsey & McBain, 1987b). The findings of Caldwell and Peplau (1982) are reflective of a common theme that runs through the research:  22  Where the sexes differed was in the nature of their interactions with friends. Women showed a greater interest in emotional sharing. Women preferred talking to activities, and, on several measures, women indicated spending more time talking to a best friend and revealing more about their feelings, problems, and personal relationships. In contrast, men appeared more interested in shared activities. Men preferred activities to talking; they more often got together with a best friend to engage in a particular activity such as a sport; and their conversations with a best friend more often centered on shared activities and interests we found that emotional sharing played a lesser role in men's than women's best same-sex friendships, (p. 731) It should be noted here that these findings of gender differences in same-gender relationships do not mean that men are not capable of being intimate or that women are not capable of engaging in sports activities together; the findings simply indicate that there are consistent modal differences in the friendships of women and men (Sherrod, 1989). However, as discussed by Surrey (1985), while gender differences such as those described above "may at times be quite subtle, and individually and culturally relative, they may represent a difference that results in enormous consequences in areas of critical human interactions" (p. 7). The investigator suggests that continuing gender differences in relationship styles may have implications for the integration or non-integration of men and women into cross-gender networks in the workplace. First, same-gender relationships may be sought in preference to cross-gender relationships for the unique qualities and sense of shared experience that they offer. While not necessarily meaning to exclude members of the other gender from organizational friendship ties, same-gender persons may do so simply because of preferred homosocial patterns of interaction. Second, the level of intimacy experienced in women's same-gender relationships may cause serious difficulties in work relationships (Woolsey & McBain, 1987a). The findings of Briles' (1987) survey of unethical behaviour in business highlight the potential difficulties. She found that when men committed unethical acts, they seemed to act from ego motivations more than women did, and that when women engaged in unethical behaviour, their actions seemed to be more related to envy and retaliations for personal slights. Briles (1987) concluded that the "issue of women's sabotage of each other often stems from a basic confusion between professional and personal relationships in business" (p. 269).  The preceding discussion has focused on actual same-gender relationships. Homosociality also includes a symbolic or ideological level of bonding which is reflected in individual attitudes and actions in regard to persons of one's own gender in general. At this ideological level, most women in the nineteenth century, for example, would have adhered to the ideology of women's sphere as the realm of the home and "the heart". Relationships with other women would have been limited to activities within this realm. Melder (1977) notes the beginnings of a sisterhood of American women who protested this limited sphere as early as the year 1800. The feminist movement has continued to this day, bringing with it changes in gender roles and relationship patterns for both women and men.  Some of the pertinent issues related to these changes are discussed in the next sections.  4. Homosocial Attitudes Homosocial attitudes refer to an "individual's affect, behaviour and cognitions regarding the same sex in general" (Woolsey, 1987a, p. 117). These attitudes have not been well defined in the literature or by research undertaken to date. However, theory and research related as closely as possible to homosocial attitudes are discussed in this section to provide a theoretical basis for the measurement of homosocial attitudes in the organizational context. The current discussion includes descriptions of the subscales of the Bonds Between Women/Bonds Between Men Scale (Bonds Scale) because the structure of homosocial attitudes that was established by factor analyses of the items contained in this scale (Woolsey et al., 1988a) provides one of the most comprehensive descriptions of homosocial attitudes developed to date. The work of other theorists and researchers is interspersed with the descriptions and conceptual analysis of the Bonds subscales to complete the discussion of the social context in which homosocial attitudes have developed. The factor-analytic work of Woolsey et al. (1988a) has established a three-part structure of homosocial attitudes for men and a five-part structure for women. (A more complete description of the development of the Bonds Scale is contained in Chapter III [Method].) The five separate aspects of homosocial attitudes, each measured by a separate subscale of the Bonds Scale, are: CrossGender Preference, Same-Gender Preference, Valuing, Jealousy (women only), and Identification  24  (women only), the current discussion begins with a brief description of the subscales common to men and women:  Cross-Gender Preference. Cross-gender persons are preferred or valued over samegender persons in certain situations, usually in relationships based on power or status differences. For example, a person with a high level of Cross-Gender Preference would prefer that persons of higher status or authority be of the other gender.  Same-Gender Preference. Same-gender persons are preferred or valued over cross-gender persons in certain situations, usually in relationships based on equality or similarity. For example, a person with a high level of Same-Gender Preference would prefer that his or her peers (e.g., coworkers, teammates) be of his or her own gender.  Valuing. Relationships with same-gender persons are valued in and of themselves because they contribute something important to the individual's life.  These Cross-Gender Preference, Same-Gender Preference, and Valuing subscales are based, most broadly, on Lipman-Blumen's (1976) definition of homosociality as "the seeking, enjoyment, and/or preference for the company of the same sex" (p. 16). The attitudes measured by these subscales arise from the differential socialization practices and reward structures for women and men. As for the theoretical underpinnings of these homosocial attitudes, two interconnected threads appear to be woven throughout the literature and research. The first thread is that gender roles have had a profound impact on the social bonding patterns of women and men. Traditionally, women and men have been socialized to enact different roles and, hence, to operate in predominantly separate spheres: for women, the nurturant, expressive domain of relationships and the home, and for men, the active, competitive domain of work, sports, and political activity.  25 Significant changes in gender roles and expectations have occurred since the resurgence of the women's movement in the late 1960s. Until that time, the spheres of men and women generally overlapped only for the purpose of heterosexual pair-bonding and procreation; otherwise, men associated primarily with other men and women related predominantly to other women as they engaged in the activities sanctioned by their gender roles (Bernard, 1981; Cott, 1977; LipmanBlumen, 1976; Melder, 1977; Pleck, 1975; Tiger, 1969). Homosociality, then, has been a prevalent social norm that has arisen from the different life experiences of women and men. The second thread is that the devaluation of women and of traits considered to be feminine has served to exclude women from men's networks (Lipman-Blumen, 1976; Pleck, 1975; Tiger, 1969), to decrease the quality and valuing of relationships between women (Caplan, 1981,1985a, 1985b; Toews, 1973; Woolsey-Toews, 1975), and to limit the level of intimacy that men are able to attain, particularly in their relationships with other men (Lewis, 1978; Pleck, 1975). Aspects of these two closely connected threads are expanded upon in tandem in the discussion which follows. Tiger (1969) and Lipman-Blumen (1976) suggest that male same-gender bonding and female cross-gender bonding are based upon an attraction to power and status, the end result of which is a hierarchical bonding pattern among dominant males which excludes women, children, and non-dominant men. Under this system, women do not have direct access to power and resources; they must derive their power indirectly through relationships with men (Miller, 1976; Lipman-Blumen, 1976; Tiger, 1969), while men, in order to maintain status and power, must relate primarily to other men in a manner consistent with the prevailing "standards of the male homosocial world" (Lipman-Blumen, 1976, p. 24). Miller (1976) contends that, because of these structural constraints on women's power, women often have been in the position of having to compete with each other for relationships with men in order to acquire resources and power, rather than assisting each other toward empowerment. In the traditional style of male bonding, men have aggregated in ways which have supported male gender role performance (Pleck, 1975), meaning that, when together, men have assisted each other in maintaining a sense of masculinity, dominance, physical and intellectual prowess, and  26 emotional control (Lewis, 1978; Pleck, 1975). Men who have been unable or unwilling to adhere to the strict rules of the male homosocial world generally have been marginalized and excluded from the traditional bonds of the male world (Lipman-Blumen, 1976). Indeed, the expression of vulnerability by men has been viewed as a betrayal of masculinity, with fears of homosexuality being triggered by incidents of emotional intimacy between men (Lewis, 1978; Pleck, 1975). In contrast, according to Bernard (1976,1981) and Miller (1976), the basis of female samegender bonding is affiliation and attachment. Furthermore, Safilios-Rothschild's (1981) summary of some research findings suggests that women play an emotionally sustaining, supportive role in their friendships with both women and men; this intimacy and support is not reciprocated by men. Men, not socialized toward learning the skills necessary to maintain intimate relationships, have tended to seek intimacy in their cross-gender relationships with women (Pleck, 1975; Safilios-Rothschild, 1981). There is some research evidence indicating that both women and men tend to prefer women as associates in intimate personal relationships, in circumstances in which personal problems are discussed and confidences revealed (Toews, 1973; Woolsey-Toews, 1975). Similar findings were noted in Rose's (1985b) study of same- and cross-gender friendships: both women and men reported receiving more acceptance and intimacy from women than from men. Woolsey-Toews (1975; Toews, 1973) conducted the first known empirical investigation of the concept of attitudinal homosociality. At the time of her initial study, she called this concept "same-sex affiliation or group belongingness" (Woolsey-Toews, 1975, p. 27); in subsequent writing (Woolsey, 1987a, 1987b), she has adopted the term homosociality. The purpose of her study was to investigate how sex-role stereotypic beliefs relate to same-gender affiliation, with her primary emphasis being on the same-gender affiliation of women. Because of the foundational nature of this study, it is described in detail. Two hundred and ninety-six female and male university students, the majority of whom were single, completed the short form of the Sex-Role Stereotype Questionnaire (Broverman, Broverman, Clarkson, Rosenkrantz, & Vogel, 1970; Clarkson, Vogel, Broverman, Broverman, Rosenkrantz, 1970; Rosenkrantz, Vogel, Bee, Broverman, & Broverman, 1968) and the Affiliation Questionnaire (Toews,  27 1973), developed for the purpose of Toews' (1973) study. The Sex-Role Questionnaire was used to assess the extent to which positively and negatively stereotyped feminine (warmth and expressiveness) and masculine (competency) qualities were incorporated into respondents' self descriptions. The Affiliation Questionnaire tapped 10 aspects of same-gender affiliation. Five factors assessed attitudes toward relationships with same- and cross-gender persons in five different spheres of activity: casual social activities, work, dependency relationships (i.e., situations in which one must depend upon a leader or a person of higher status), important tasks or crisis situations, and personal friendship. The remaining five factors measured pride in, loyalty to, and attitudes toward one's own gender as a group, and willingness to accept sex-role labels and to be described as a typical person of one's own gender. Toews' (1973) results indicated that men and women had very different styles of samegender bonding; on 9 of the 10 factors, the results for men and women were significantly different. For example, on the factor measuring acceptance of sex-role labels, the women indicated that they did not like being described as feminine; the men, on the other hand, enjoyed being seen as masculine. On the factors assessing work relationships, crisis situations or important tasks, and casual socializing, both men and women preferred men as companions. In other words, everyone rated men as preferred "playmates, workmates and as better people to have around when there's trouble or when you need a leader" (Woolsey-Toews, 1975, p. 29). For women, there was a significant interaction between personally held sex-role stereotypes and their degree of same-gender affiliation. The more competent and, hence, less stereotypically feminine a woman reported herself to be, the less likely it was that she would evidence pride in women as a group and a sense of loyalty to other women, and the more likely it was that she would indicate a wish to disassociate herself from other women. Women who described themselves as being low in competence -- that is, the most stereotypically feminine women -- indicated the strongest sense of affiliation with other women. Being competent seemed to mean foregoing relationships with other women. The women who described themselves as the most competent were the most likely of all to express preferences for men as work associates, leaders, and  28 companions. This finding, combined with the fact that men expressed strong preferences for other men as workmates, has serious implications for women trying to advance in the work world. In this study conducted in the early 1970s, neither men nor competent women indicated a sense of bonding with women on the factors related to work and leadership. The results of more recent research (Liden, 1985; Trempe, Rigny, & Haccoun, 1985) indicate that it is the degree of upward influence held, rather than gender, that is related to satisfaction with or preference for one leader over another. Because women, until recently, have not possessed significant access to power and resources outside the domestic realm, they have not engaged in the type of resource exchange relationships with each other that are characteristic of men's same-gender relationships (LipmanBlumen, 1976). One result of the increasing entry of women into the paid labour force is that some women now are able to offer to others the instrumental support and access to resources previously available only from men. However, because women have tended to enter at and to remain in, lower level and lower paying positions, the majority of employed women have remained in the position of having much less than men to offer each other in terms of assistance with career advancement and resource acquisition. Traditionally, women seeking career advancement in male-dominated organizations have had to seek assistance from men rather than from other women. For example, Hennig and Jardim (1977) studied the life and career histories of 25 women who had begun careers in business and industry around the time of the Great Depression and had reached top management positions by 1970. They found that all of these women had male mentors who were key influences on the success of their careers. These 25 women focused on the development of competence and on career advancement for the first 10 years of their careers. They had little time for personal relationships other than the key relationship with their male mentor. In contrast, and as suggested by Kanter (1977), women located in dead-end jobs may substitute social recognition for the rewards of career advancement. Kanter (1977) contends that "it is low-opportunity people who seem to find their greatest satisfactions at work through connections with others" (p. 147). Hence, women who perceive little opportunity for career advancement may focus even more on personal relationships  with other women as a major source of job satisfaction rather than attempting to achieve job satisfaction through successfully attaining desirable promotions and assisting other women to do the same. The development of resource exchange relationships between women has been both assisted and attenuated by feminist reactions to restrictive and devalued female gender roles. The traditional female gender role and relationships between women were supported by the structure and norms of the strong "female world" (Bernard, 1976,1981) of the nineteenth century, a world which has eroded in this century. Since the origins of the feminist movement in the early 1800s (Melder, 1977), the active reevaluation of women's position within society by women who have felt severely restricted by the female gender role has precipitated both a new sense of bonding or sisterhood among feminist women and a series of powerful divisions among women of different persuasions in regard to female gender roles (Caplan, 1981,1985a, 1985b; Melder, 1977; Rawlings & Carter, 1977; Woolsey & McBain, 1987a). As stated by Nancy Sahli (1978):  In a sense, the feminist movement itself subverted the heightened emotional commitment which had typified women's relationships during most of the nineteenth century. As women began to be perceived by themselves and others as being capable of rational, intellectual thought, it seems evident that they would want to use this ability, rather than their emotions, to make decisions advancing their position vis a vis the male world, and, in their search for equality with men, that they would perceive this capacity as being on a higher status scale than that of the emotions, (p. 26) Some women, then, in their attempts to operate on an equal status with men, have devalued traditional feminine qualities. For example, some women in higher professional and organizational ranks have been found to identify more strongly with the masculine aspects of their personalities (Fagenson, 1985; Hennig & Jardim, 1977) and to reject the traditional female qualities (Dellacava & Engel, 1979). This rejection or devaluation of feminine traits has been shown to result in a decreased sense of bonding with other women, both in actual relationships and in general attitudes toward other women (Dellacava & Engel, 1979; Marshall, 1984; Toews, 1973; Woolsey-Toews, I975). Dellacava and Engel (1979), Fagenson (1985), and Laws (1975) interpret this phenomenon in terms of the structural constraints placed on women in high-level professional careers. Dellacava and  30 Engel (1979) suggest that in order to be successful in a male-dominated career, women must identify strongly with masculine norms and reject the feminine: Professional women need to see themselves as special and unique. Women in our society are socialized with the expectation that they will be wives and mothers and that those roles ought to be the key statuses in their identity structure. Consequently, professional women have an internal conflict, and women who opt for the traditional roles represent an important [but rejected] aspect of their [professional women's] own identities. To remain committed to the professional role means that one must be and remain estranged from (marginal to) other women. The latter must be a negative reference group, they must be rejected. In the baldest terms: given the present society and its culture, the professional woman has no choice but to remain uninvolved with other women and their concerns, (p. 508) Kanter (1981) suggests that the rejection of other women by upwardly mobile professional women is a result of upper-level women's token or minority status in their work groups. She contends that "to the extent that tokens accept their exceptional status, dissociate themselves from others of their category, and turn against them, tokens may be denying parts of themselves and engaging in self-hate" (p. 83), thus producing considerable inner tension even though their work performance may be successful. This phenomenon of strongly identifying with masculine norms and rejecting the feminine does not appear to be true for all professional women in male-dominated careers. McBain and Woolsey (1986) studied the career and life role aspirations (career, home and family, and personal) of 29 women in their final year of a Bachelor of Commerce programme. While their most preferred roles for the projected five-year period following their graduation were professional roles, they did not indicate any rejection of the more traditional roles of wife, mother, and homemaker. These results are supported by Yogev (1983) who stated that: Data on career aspirations imply that the career-marriage conflict is rapidly diminishing in importance. Its persistence as an issue for discussion is more a function of educators and counselors than the perceptions of women themselves, especially of young women, (p. 224) As more and more women enter the paid work force and become an increasing presence in previously male-dominated areas, the pressure for professional women to dissociate from other  31 women would appear to be diminishing. The remaining two homosocial attitudes, as measured by the Bonds Between Women Scale (Woolsey et al., 1988a), represent two aspects of the struggle to develop supportive, affirming relationships between women who, as members of a "minority" group, have special tensions in their relationships that are not seen in relationships between "majority" group members (men). The two subscales are:  Jealousy. Jealousy and other such negative feelings between women are a manifestation of the in-group aggression seen in other minority groups. Jealousy between women precludes or breaks down in-group ties and mutual support between women.  Identification. Women identify with other women as members of a disadvantaged social group, in a pattern similar to that observed in other minority groups (e.g., ethnic or racial minorities). This identification may be expressed in various ways: pride in the accomplishments of other women, joining in collective efforts to improve the position of women in society, and making a special effort to help other women as individuals.  Rawlings and Carter's (1977) adaptation of Allport's (1955) minority group theory provides the main theoretical framework for the homosocial attitudes of Identification and Jealousy. Rawlings and Carter contend that while all women may not experience or perceive themselves to be the recipients of prejudice or discrimination, low social status is accorded to women as a group. Women's low social status is a result of prevalent negative images of women in the culture as a whole and of barriers to women's full participation in social, political, and economic spheres. To lessen the effect of this cultural devaluation, women have developed ego defenses similar to those of other victims of discrimination. These ego defenses can be either "intropunitive" or "extropunitive" in form; that is, the anger that results from being a victim of discrimination can be turned inward to one's own group or outward to the dominant group.  32 Intropunitive ego defenses include denial of membership in one's own group (i.e., lack of Identification) and in-group aggression such as competing with other women, back-biting or gossiping about other women, and putting down other women. The Jealousy subscale represents this type of in-group aggression among women. Intropunitive defenses serve to erode in-group ties because the anger which results from unjust treatment is directed toward other women. The extropunitive defense of strengthening in-group ties (e.g., by establishing support groups for women, engaging in cooperative political action) is, in some ways, the converse of denial of membership. As defined, strengthening in-group ties represents the presence of Identification with other women. The presence of this extropunitive ego defense would tend to enhance the likelihood of supportive relationships between women. The dynamic of either identifying with or dissociating from members of one's own group was described, in part, in the preceding presentation of theoretical framework for Cross-Gender Preference, Same-Gender Preference, and Valuing. The following study by Marshall (1984) describes how women can move from a position of denying their own femaleness and avoiding relationships with women to a position of identifying with other women. Marshall (1984) conducted interviews with 30 women managers, all but four of whom lived in or around London, England. Half were managers in book publishing companies and half in retailing companies. They ranged in age from 26 to 50. Four respondents in the publishing industry were directors, four were in senior management positions, and seven were in middle management positions. Two of the respondents in the retail industry were directors, eight were senior management, and five were middle management. From Marshall's description of the difficulty she had in locating suitable women for her sample, it can be surmised that her sample was drawn from skewed groups (15% or less women) (Kanter, 1977). The purpose of Marshall's interviews was to gain an understanding of the women's lives as wholes, with a focus on their career development. Only a few of Marshall's findings -- those related to the issue of managing the awareness of being female ina male-dominated environment ~ are reported here.  33  Marshall (1984) divided her sample into three groups, according to their method of managing their awareness of being female. One group, the majority of the respondents, were strongly opposed to identifying themselves as disadvantaged because of their gender. Most of these women viewed the Women's Movement as "complaining too much"; they stated that being a woman was irrelevant. The women in this group often contradicted themselves with passionate discussions of behaviour or attitudes they had adapted to counteract the disadvantages of being female. The main tactic used by this group of women was to dull their awareness of being women. They identified themselves as particular cases separate from women as a group. They engaged in private strategies for survival and career achievement and, being successful at this, saw no reason to identify with other women as a group. Some of these women chose to minimize their interactions with other women at work, in order to decrease the chances of being identified with other women. This group of women might be expected to receive low scores on the Identification and SameGender Preference scales of the Bonds Between Women Scale. Another group (three women) expressed painful awareness of the disadvantages of being female in a male-dominated world. All three of these women had been prompted by recent circumstances to become acutely aware of themselves as women, thus shattering the previous denial which had maintained a relative calm. All three were dealing with issues of identity and their attitudes and feelings were shifting dramatically as they developed new perspectives. None had yet developed new perspectives or values which enabled them to be comfortable with their awareness of gender. A third group, representing about a third of the total sample, had already established a clear and comfortable sense of themselves as women. Most of these women had experienced prejudice and discrimination because they were women and were very aware of the inequalities in reward between themselves and their male colleagues. Marshall (1984) summarized their position with this statement: These managers showed greater concern about the position of women as a group, and of themselves as women, than did the majority of interviewees. They brought analyses along these lines to their answers, but seemed personally untroubled by  34 them. Some translated these attitudes into an advocacy of women's rights. Others took up the possibility of discrimination against them as a challenging gauntlet. Several of the latter expressed an edge of bitterness or resentment, but this was typically balanced by cynicism and self interest. 'But I wouldn t change it, its [sic] exciting and challenging' was a typical comment, (p. 153)  These women might be expected to receive high scores on the Identification scale of the Bonds Between Women Scale and to report more women in their networks than the first group of women. Three publications (Briles, 1987; Eichenbaum & Orbach, 1988; Woolsey & McBain, 1987a) have documented the potential destructfveness of work relationships between women when they are characterized by high levels of jealousy. The results of Briles' (1987) survey of unethical behaviour in business indicated that when women commit unethical acts, they "more often act out of jealousy, envy, or fear; whereas men tend to act more from ego motivations" (p. 20). The women in Briles' (1987) study reported a different character to the unethical behaviour they had experienced from men and women:  When the women in our study reported unethical behavior by men, it usually revolved around issues of sexual harassment, discrimination in employment, and failure to be promoted. But their reports of unethical behavior by other women concerned issues of power and competition, reactions to personal slights and envy. (Briles, 1987, p. 88) Unethical acts between women were characterized by vengefulness, secretiveness, manipulation, back-stabbing, and a sense of character assassination. Eichenbaum and Orbach (1988), writing from the personal and clinical experience they had gained in working with women and in women's groups since the 1970s, state that "envy is among the most painful feelings women experience toward other women" (Eichenbaum & Orbach, 1988, p. 92). Over the past ten years, as gender roles and women's expectations have changed, they have observed an increased incidence of envy between women, so much so that it is "threatening to disrupt even the best of female friendships" (Eichenbaum & Orbach, 1988, p. 92). They describe the following dynamics:  35 One woman's success is often threatening to another. But why? The answer has been clearly expressed by hundreds of women we have worked with. They describe feeling deserted, left, abandoned as the other woman moves on and develops. It feels as though the successful one is turning her back, leaving her friend to stay stuck in the space they once shared. At the same time the woman who is a success feels alone in the new and unknown space. (Eichenbaum & Orbach, 1988, pp. 94-95)  Woolsey and McBain's (1987a) clinical observations and analyses led them to a similar conclusion: the professional success or personal power of one woman may give rise to feelings of jealousy, rage, and powerlessness in another woman who sees no direct route toward success for herself. If not handled carefully, the resulting conflict between the two women can have devastating consequences for their relationship and for their productivity at work. From these descriptions, it is clear that high levels of jealousy and threat in relationships between women preclude the development and maintenance of supportive work relationships between women. Writers such as Forisha and Goldman (1981) suggest that women, rather than rejecting each other, can be of invaluable support and practical assistance to one another "as they move out of traditional female systems and into those dominated by males" (p. 149). They contend that because women can rarely become a part of male support networks in organizations, they must establish a parallel support system of their own. Establishing a women's support network means valuing and preferring other women enough to identify with them and to assist them in career advancement, rather than focusing exclusively on relationships with men (cross-gender preference) in attempts to achieve career success. Jealousy between women is an issue that must be addressed and worked through if women are to develop consistently supportive relationships with one another.  5. The Relationship Between Homosocial Attitudes and Behaviour One of the main purposes of the present study was to assess the relationship between homosocial attitudes and specific behaviours (degree of relationship formation with same- and cross-gender persons, as indicated by the number of persons listed as network members; the level of mentoring behaviours received; and the levels of relationship provisions received). At this point, it  36 should be noted that, in the current study, behaviours were not directly observed. Rather, respondents' perceptions of behaviours served as the measure of the behaviours. Nevertheless, it seems important to discuss some issues concerning the relationship between homosocial attitudes and behaviour. Baron and Byrne's (1987) discussion of attitudes, and their summary of research that has examined the relationship between attitudes and behaviour, formed the theoretical basis for the current discussion of the relationship between homosocial attitudes and behaviour. All references in this subsection to the Bonds Between Women/Bonds Between Men Scale (Woolsey et al., 1988a) are the extrapolations of the current writer, not of Baron and Byrne. Baron and Byrne (1987) define attitudes as "lasting, general evaluations of people (including oneself), objects, or issues" (p. 116)- lasting, in that attitudes tend to persist over time; and general, in that the evaluation tends to involve some degree of abstraction. Consistent with this definition, the Bonds Between Women/Bonds Between Men Scale (Woolsey et al., 1988a) was designed to measure attitudes toward same-gender persons and relationships in general, attitudes which are believed, but not yet proven, to persist over time. Baron and Byrne (1987) state that most psychologists believe that attitudes have three components - affect, behaviour, and cognition: The affect component refers to positive or negative emotions - our gut-level feelings about something. The behavior component involves our intentions to act in certain ways, to engage in behaviors that are somehow relevant to our attitudes. Finally, the cognition component refers to the thinking and interpreting that goes into forming or using an attitude. Each attitude, then, is made up of a cluster of feelings, likes and dislikes, behavioral intentions, thoughts, and ideas, (p. 116) The Bonds Between Women/Bonds Between Men Scale (Bonds Scale) captures aspects of all three of these dimensions. The Bonds Scale is composed of three subscales (for men) and five subscales (for women) which measure different aspects of homosocial attitudes. Each subscale is composed of a cluster of items, each of which represents a cognition, behavioural intention, preference, like or dislike, or feeling about same-gender relationships or persons in general.  37 Baron and Byrne (1987) cite Wicker's (1969) review of studies that examined the link between attitudes and behaviour. Wicker found that, in study after study, attitudes and behaviour were weakly or negligibly correlated. Since the time of Wicker's (1969) review, however, new evidence has indicated that, under certain conditions, attitudes can predict behaviour accurately. Baron and Byrne (1987) list several important factors which determine the strength of the relationship between attitudes and behaviour. The first factor discussed by Baron and Byrne is attitude specificity. Citing the review of research conducted by Ajzen and Fishbein (1977), Baron and Byrne suggest that "to predict overt behavior from attitudes, it is usually more effective to look at specific, narrow, and precise attitudes instead of general or global ones" (p. 141). Ajzen and Fishbein found that studies measuring global attitudes showed weak or negligible correlations with behaviour, whereas studies examining specific attitudes showed strong correlations with behaviour. A visual inspection of the Bonds subscales and items reveals that the Cross-Gender Preference and Same-Gender Preference subscales are the most specific of the five subscales; therefore, the strongest relationships between attitude and behaviour would be expected for these subscales. The other three subscales (Valuing, Jealousy, and Identification) are also quite specific, although not quite as specific as the first two scales mentioned. Therefore, the relationships between these three subscales and behaviour might be expected to be somewhat less strong than for the first two subscales. The second factor listed by Baron and Byrne as affecting the strength of the attitudebehaviour relationship is the strength and accessibility of the attitude. Strong attitudes - those that are more accessible or more readily brought to mind - have been found to predict behaviour more effectively than weak ones. Two factors are related to attitude strength. First, "attitudes formed by direct personal experience tend to be stronger and tend to predict behavior better than other attitudes [e.g., those passively acquired by observation]" (Baron & Byrne, 1987, p. 141). Creating an example that would be relevant to the current study, one might infer (from Baron and Byrne's work) that persons who have had direct personal experiences with both women and men in the workplace have had more opportunities to form experienced-based as opposed to observation-based attitudes  38 than persons whose work contacts have been limited to one gender. Baron and Byrne suggest that having a vested interest in the issue in question also increases the strength of the relationship between attitudes and behaviour. Again, to create an example relevant to the present study, one might say that if relationships with both women and men are perceived as important to career advancement, then attitudes toward these relationships will have direct personal relevance. In summary, the more often a person has "thought about a particular attitude, the more likely it is to come up again and to influence... behavior" (Baron & Byrne, 1987, p. 142). Citing the work of Fazio, Powell, and Herr (1983), Baron and Byrne (1987) contend that:  If accessibility is high, then, even general attitudes can exert a strong influence on behavior. This has to occur through a three-step process. First, something calls the general attitude to mind. Second, the general attitude influences how the person perceives the situation, "coloring" judgements and interpretations. The general attitude operates like a schema in creating expectations and guiding attention, encoding, and retrieval. Third, behavior is determined by these judgements and interpretations of the immediate situation.... If the general attitude is never accessed, it won t affect behavior, (p. 142)  B. CROSS-GENDER RELATIONSHIPS Changing gender roles and women's increasing entry into the labour market are precipitating changes in cross-gender relationships as well as in same-gender relationships (Sherrod, 1987). Colwill (1982) contends that, in the modern workplace, women and men are meeting on common turf and a new kind of partnership is being forged between them, a partnership based on the ability of all human beings to be masculine (to be assertive, to assume leadership, to make decisions) and to be feminine (able to be sensitive, emotionally expressive, and considerate of the needs of others). This is a time of transition in which new rules and norms for cross-gender relationships in the workplace are evolving (Safilios-Rothschild, 1981). As stated by Colwill (1982):  One of the adjustments that many husbands and wives have made in the past decade and that many more will make in the next decade is the waiving of exclusive rights to the role of opposite-sex friend of the spouse. While sexual exclusivity will probably continue to be a marital ideal in the foreseeable future, it is inconceivable that men and women will work together as peers and not form social ties ranging from casual acquaintance to intimate friendship, (p. 49)  39 The changes envisaged by Colwill are in direct contrast to the separation between the male and female worlds in the nineteenth century. Smith-Rosenberg (1975) noted that in the letters and diaries of young women, references to women and men were characteristically different. Relationships with other women were characterized by emotional intimacy and spontaneity while relationships with men frequently lacked these qualities. Men were represented as "an other or out group, segregated into different schools, supported by their own male network of friends and kin, socialized to different behavior, and coached to a proper formality in courtship behavior" (pp. 20-21). While young women sought marriage and domesticity, in their writings they referred to young men as distant and held at bay. Cross-gender relationships were very formal. While most cross-gender relationships today are not characterized by this degree of formality, social ties between women and men can pose dilemmas. Clawson and Kram (1984) state that: In most cross-gender cases nowadays, the coach or mentor is a male and the protege is a female, but more and more women in management are also having to ask themselves the question that men ask: How do I manage developmental relationships with subordinates of the other sex and at the same time manage the possible pitfalls associated with close male-female working relationships? (p. 23) Kram's (1985) findings in her study of mentoring relationships indicated that the friendship aspect of mentoring is more limited in cross-gender relationships than in same-gender ones. Men and women avoided interactions in informal settings because of concerns that these interactions would be interpreted sexually both by the parties involved and by other organization members. As a consequence, important collegial interactions, which so often occur in informal settings, were missed. Some female executives in Reich's (1986) study also noted difficulties in cross-gender relationships. They commented that mentoring relationships with males sometimes became sexual, and that men's offers to assist women were sometimes sexual advances in disguise. Concerns about keeping the level of intimacy in cross-gender relationships to an appropriate level can preclude the closeness which enables a coach or support person to have a greater effect on the subordinate's learning (Clawson, 1980).  40  C. HOMOSOCIALITY AND ORGANIZATIONAL NETWORKS Over the past 15 years, a plethora of writers in both the popular and academic literature have discussed the negative effects on women's career development of being excluded from men's networks (e.g., Abramson, 1979; British Columbia, MISBD, 1979; Forisha & Goldman, 1981; Hennig & Jardim, 1977; Josefowitz, 1980; Kanter, 1977; Lincoln & Miller, 1979; Lovelady-Dawson, 1981; Riger & Galligan, 1980; Viega, 1976). During this same period, women's networks have sprung up across Canada (Fraser & Hendry, 1982) to provide women with "Psychic Turf" (Steinem, 1982) -- a place of validation and information-sharing among women. As summarized by Black (1982): Groups run for and by women are critical. They are our "Psychic Turf', our place to discover who we are or could become. Network participation will allow us to become whole persons. They provide a time out from our roles as mothers, wives, and workers and help us to develop the skills necessary for our work self, that part of self which has traditionally been assigned to the males of our culture, (p. 12) The theory behind the establishment of women's networks in the workplace is that: Women, academic women and other professionals, tend to be unaware of the unspoken expectations of the male work world, even the most competent and highly-trained women. Seldom do they have the opportunity to become part of the informal information networks that are such an important part of any organization. (Power, 1981, p. vii) The creation of women's networks is designed to address this deficit, to help women to actualize their full potential academically or professionally by providing an alternate source of information and support. While in theory the establishment of women's networks should assist women's career development and add to their overall job satisfaction, there are two factors that work against the formation of effective supportive relationships between women. First, certain patterns of homosocial attitudes may preclude the development of supportive relationships. For example, a woman with the following pattern of homosocial attitudes is unlikely to develop supportive work relationships with other women: high levels of Jealousy and Cross-Gender Preference, and low levels of Same-Gender  41 Preference, Valuing, and Identification. This woman will be much more likely to develop relationships with men in her workplace and to exclude women from her relationship constellation. Second, traditional patterns of female homosociality at a behavioural level (i.e., women's proclivity to form intimate relationships, regardless of the setting) may cause difficulties in work-based relationships between women. While aspects of both of these factors have been developed throughout this chapter, further discussion of the work context in which women's relationships are to develop, or not to develop, is warranted.  1. Independent Achievement Versus Development In Relationship The work of Hennig and Jardim (1977) provides some insight into the difficulty facing women in their integration into male-dominated organizations. In 1973, they began in-depth interviews with 45 women in senior management positions in a large public utility and 63 women in management positions in the banking industry. Their ages ranged from 27 to 58 and all but one had worked continuously since finishing high school or college. A significant theme that emerged from the analysis of these interviews was the reliance of these women on individual competence and self-improvement. They believed that their superior performance alone would result in their being chosen for advancement. This belief was founded on the premise that promotions would occur according to the formal system of roles, policies, and definitions put forth by the organization. They made no mention of "the informal system of relationships and information sharing, ties of loyalty and of dependence, of favors granted and owed, of mutual benefit, of protection - which men unfailingly and invariably take into account to however greater or lesser a degree" (Hennig & Jardim, 1977, p. 31). Unaware of the important function of this informal system, individual women were left disappointed and immobilized when the formal system failed to reward them with career advancement. As a contrast to the information gained in the interviews with these 108 women, Hennig and Jardim (1977) examined the life and career histories of 25 women who had begun their careers around the time of the Great Depression, and who had reached top management positions in  42 American business and industry by 1970. They found that the assumptions and attitudes shared by these 25 women were in marked contrast to those of the previous 108 women. These 25 women began their careers working for lower-level male executives in manufacturing, retailing, banking, insurance, finance, consumer goods, chemical industry, or consulting firms. A few of the 25 changed jobs within the first two years. Then, for the next 30 years, all were employed by the same firm. They established close relationships with their male bosses who supported, encouraged, taught, and backed them in dealings with male customers and clients and other men in the company. The women were sustained by these relationships:  The boss acted as sales agent for the woman wherever he sent her, both inside and outside of the company. He used his reputation to develop hers, and his respect from others to gain acceptance for her. In times of direct confrontation with any group or individual, he would act as a buffer and place himself between the woman and her opponent. He was the protector and she the protected. His support helped provide her with the extra confidence she needed to take on new responsibilities, new tests of her competence and new positions. He reinforced her own emphasis on competence as an issue of paramount importance. (Hennig & Jardim, 1977, p. 157)  These relationships provided the women with the foundation they needed to gain new and broader skills. And acquire skills they did, each becoming more competent at her own job, the job below her, and the job above her than any man available. This level of competence, however, was attained at great personal cost: for the first 10 years of their careers, they had little time for personal relationships, visiting family, or vacations. Most remained single until at least age 35, few maintained a long-term relationship with a man outside work, and the majority had only one or two close women friends (most frequently these were sisters or other relatives). Relationships with their male peers were few and tended to be direct, factual, work-centred, and emotionally distant, a strategy employed to avoid being accused of acting like a woman. The deep and long-lasting friendship with their boss was the one exception to this pattern. During the first 10 years of their careers, each woman began as the secretary or administrative assistant to one man and moved upward in the organization with him, at his request, as he advanced. By 1970, each of these women managed to be successful in a male-dominated system  43 by putting in extraordinary amounts of personal effort to achieve a high level of competence and by aligning herself with a man with enough power in the organization to help her navigate herself through it. Each woman could not have made it without the strong support of her male ally.  2. The Psychology of Tokenism The pattern exemplified by the 25 successful women in Hennig and Jardim's (1977) study has been defined as tokenism by Laws (1975). Laws contends that tokenism, "defined as a form of interclass mobility.... is likely to be found wherever a dominant group is under pressure to share privilege, power, or other desirable commodities with a group which is excluded" (p. 51). A member of an excluded group is granted mobility into the dominant group's domain via a significant relationship with a member of the dominant group. However, the quality and quantity of mobility is severely restricted. The Token, as a member of a numerically underrepresented group, functions on the dominant group's turf only under license from the dominant group. The Token remains in a marginal position, never becoming an equal member of the dominant group. The Token and her Sponsor develop a special relationship in terms of affection, frequency of interaction, and complementary exchanges. The Token and Sponsor provide confirmation for each other. The Sponsor vouches for the Token, shapes her role performance, and exerts social control over the behaviours of other members of the dominant group toward her. Laws (1975) presents tokenism within the gender-class system which assigns to men, as a class, higher status than women: Women's differential access to opportunities and rewards makes them a disadvantaged group, in spite of their numerical dominance. Different evaluation contributes to an image of women as inferior. And the expectation of different attributes operates within a framework in which male is normal, and different is deviant. Thus tokenism must include techniques for neutralizing or managing deviance, (p. 53)  Women, then, are defined as deviant within a male-dominated system. However, it is common in both classes (men and women) to deny the effects of gender on opportunity structures, aspiration, and achievement. Laws (1975) contends that "the cultural denial of gender classes and  44 the disesteem in which women are held both have profound consequences for the psychology of the Token" (p.53). The Token is not likely to characterize women in any particular way. If she does, she is highly unlikely to classify herself according to the traits of her gender as a whole. On the other hand, the Token's gender is likely to be salient to others, despite her denial of the effects of gender. She is assigned to primary-deviant status simply by virtue of being born female. Then, because she aspires to the privileges and attributes of the dominant class, she becomes doubly deviant: she is a woman and she is refusing the constraints of this ascribed status. The double deviant woman values in men the attributes associated with men and invidiously devalues the attributes associated with women. She prides herself in her ability to perform according to the valued qualities of the dominant group (e.g., rationality, coolness, critical thinking). The double deviant views herself primarily as an individual who has more in common with men than with women. It is her chosen path of upward mobility that determines that she have these perceptions; if her career is to advance, she must resemble and identify more with the dominant group than with other women who are not upwardly mobile. Laws (1975) emphasizes that not all double deviants become Tokens in a male-dominated system. The double deviant is recruited and moves through a series of stages toward the role of Token. Not all recruits, however, make it to the destination of Token. Laws (1975) speculates that there are "two aspects of differential socialization which move the double deviant out of the mass of primary deviants and put her on the track to an equivocal participation in the dominant group" (p. 54). The first aspect is a history of significant relationships with a number of sponsors. The second is a history of selectively associating with members of the dominant group (men) rather than with other women. A special talent demonstrated at a young age by the double deviant brings her to the attention of a sponsor. The child shares with this first sponsor, and with sponsors to follow, a special relationship in which she is treated as a special person and is exposed to opportunities which would normally not be accessible to her.  45  As a double deviant progresses toward her destination of Token, she is progressively trained to emulate the qualities of the dominant group. The further she progresses along this path, the more she will associate with men rather than with women, and the more she will imitate and identify with dominant class attributes. Her Sponsor is her major and sometimes only source of support: he is a lifeline, a vital source of validation, esteem, and acceptance. She is likely to be isolated from other women because of her choice to identify and associate with men and to devalue women. Token women may serve as "gate-keepers" to the dominants' turf, helping to screen less exceptional outsiders (i.e., the majority of other women). Hence, a Token woman is likely to draw the wrath of women who perceive her to be blocking their advancement. Becoming a Token is a method of survival in a male-dominated system. Laws (1975) suggests that double deviants who do not fully accept the role of Token represent a large number of women who are not promoted. A woman who wants to succeed and to be upwardly mobile in a male-dominated system is in a very difficult position: Having the resource of a female support group might well lessen the Token's psychological reliance upon the Sponsor and her psychological vulnerability to the male role set. However, the female reference group can do little to lessen the Token's material vulnerability to her colleagues. Rather, any affiliation with the Women's Movement is likely to reduce their tolerance for and evaluation of her. (Laws, 1975, p. 66) Kanter (1977) has also written about tokenism. She identified four group types in organizations according to the different proportions of kinds of people in the group. People may differ in terms of important master statuses such as gender, ethnicity, or race. Uniform groups are composed of only one type of person and, hence, have a ratio of 100:0. In skewed groups, there are many more of one type than the other (ratio around 85:15). The numerically dominant types exert enough control over the group and its culture to be called "dominants". The very small number of the other type in skewed groups are called "tokens". Tokens often become symbols or representatives of their type rather than individuals. As such, they are highly visible. It is very difficult for tokens to generate a strong enough alliance with each other to exert any control in the group. In tilted groups, the proportions are less extreme. With ratios of about 65:35, the numerically  46 dominant types are called a "majority" and the tokens move to the status of a "minority". Members of the minority group are present in large enough numbers to develop allies among themselves and to form coalitions which can affect the whole group and its culture. Individuals can be both a part of the minority subgroup and individuals differentiated from the group as a whole. Finally, balanced groups occur at ratios of 60:40 to 50:50. With more equal numbers of each type, majority and minority are potential subgroups that may or may not develop their own identity. Outcomes for both types of individuals will depend more on other personal and structural factors.  3. Advancement Through the Activities of Informal Networks: Blocks and Opportunities Hennig and Jardim (1977) contend that it is the informal structure of organizations that will either block or enhance efforts toward equal opportunity for women: In most organizations the informal system of relationships finds both its origins and present function in the male culture and in the male experience. Its forms, its rules of behavior, its style of communication and its mode of relationships grow directly out of the male developmental experience. This cannot be viewed as either good or bad. It is real. Men founded and developed the vast majority of the organizations we know. Men made them places where they could work and live and their settings were intended to be both comfortable and familiar. And if organizations in general are dominated by a male culture, then we need to note that at the management level, and particularly in its higher ranks, the informal system is truly a bastion of the male life-style, (p. 13) Men, then, are the "insiders" within this system: they understand and support each other, the informal structure, and its rules; they have a common experience base of playing, competing, and learning together, and they share common aspirations. Women, attempting to enter this system with a different base of experience and style of relating, are the "outsiders" (Forisha & Goldman, 1981; Hennig & Jardim, 1977). Hennig and Jardim (1977) contend that insiders, threatened by punishment if equal opportunity laws are not adhered to, can use the informal system to discreetly block gender integration. While legislation can control formal structure, "beliefs, attitudes and assumptions which people have about themselves and each other and their resulting willingness or unwillingness to accept each other are untouched by law" (Hennig & Jardim, 1977, p. 14). Abramson's (1979)  47 accounts of gender discrimination within professional employment, particularly in education and government in the United States, clearly illustrate this phenomenon. During the late 1970s, she interviewed about 150 people: women who had filed discrimination complaints, co-workers of complainants, employers, affirmative action officers, and so forth. Her book contains descriptions of case after case in which the career advancement of women was actively blocked by the actions of their (male) superiors. Information critical to career advancement was withheld from women, while the men around them received the necessary information; performance ratings were written in such a way as to defame the women's credibility; sexual harassment by male co-workers and superiors was allowed to go virtually unchecked; and women repeatedly had been given assignments well below their ability level and position. In many of the cases she described, women were systematically demoted because of these blocking actions, to the point that their careers were in ruins. In many cases, equal employment opportunity legislation was in place and rulings of investigating bodies were in favour of the women. However, the organizations for which the women worked frequently found ways to avoid the rulings and to continue to hold the women back. Investigations often were protracted to such an extent as to be virtually useless in helping the women maintain their positions or to advance in accordance with their qualifications. In summarizing the work of the American Equal Employment Opportunity Commission in regard to individual claims of discrimination, Abramson (1979) states: There are thousands of people who have filed charges of discrimination over the years who will get no remedy or who, at best, will gain a few thousand dollars or perhaps a belated promotion for their trouble. There is no guarantee in the system that, once they have settled, they will not immediately experience retaliation. There is no guarantee, and little likelihood, that having filed a complaint, settlement will place them in a nondiscriminatory situation. Some could probably turn around following settlement and instantly file a second, equally legitimate complaint of discrimination, (p. 63) The findings of a study initiated by the Equal Employment Opportunities Committee of the Public Service Commission (British Columbia, MISBD, 1979) (discussed in greater detail below) are consistent with Abramson's (1979) accounts of blocked career advancement of women by exclusion from male organizational networks. The strategies of exclusion described by women in the public  48 service included isolating the solitary woman from work contacts, cutting off information, reorganizing work flow to reduce the woman's control of her area of responsibility, and job reorganization to reduce its scope and the credibility of the woman occupying it. The authors note that these forms of exclusion are common strategies used to limit and contain the activities of any individual, male or female. However, these tactics were widespread among the accounts of professional women. Women consistently found themselves left without meaningful work or the means by which to develop their work after a formal reorganization or a routine change in work allocations. Hennig and Jardim (1977) suggest that laws alone will not ensure the advancement of women to higher levels of management. Success as a manager and access to upward mobility also require of women "understanding and skill at working in and with the informal system of relationships in which management jobs are embedded" (pp. 14-15). They go on to say that women's lack of career advancement as compared to men has much less to do with acquiring technical knowledge than it has to do with basic gender differences in perception which have prevented women from acquiring new and very different skills in regard to managing the informal system of relationships. These different perceptions, contend Hennig and Jardim (1977), have "left women trapped in [lowlevel] supervisory positions and too often branded as lacking in management potential" (p. 15). The work of Hennig and Jardim was expanded upon in 1978, when the Equal Opportunities Committee of the British Columbia Public Service. Commission (British Columbia, MISBD, 1979) initiated an investigation of how the structure and conduct of work in the B.C. Public Service contributed to the differential location of men and women in the organization's management. (In 1977, out of 431 management positions in the Public Service, only 18 were held by women.) The study examined the advancement of men and women, focused on advancement into managerial positions from lower-level positions within the organizational hierarchy, and investigated the ability of persons within these lower-level positions to gain the prerequisite knowledge and experience for merit-based advancement into management positions.  49  Two methods of data collection were employed. First, in-depth interviews were conducted with 54 men and 80 women public servants in selected job classifications and in management positions across British Columbia. In addition, interviews were conducted with 14 key senior officials in the capital city of Victoria. Finally, a self-ad ministered questionnaire was distributed to 1200 public servants in selected job classifications. The response rate to the questionnaire was 72%. The questionnaire explored respondents' career experiences and the organization of the work in their current positions. The findings indicated that advancement is not simply a matter of producing work. Rather, advancement arises from progressive changes in the relation of an individual's work to the organization, and involves relating to the organization in a manner that is wider than the individual's current job description. The individual who wishes to advance must find ways to direct increasingly different activities, to exercise more decision-making power, and to control an increasing number of resources in a manner consistent with the organization's goals. The skills of advancing one's work in this way are taught within the everyday working environment in the informal network of relationships embedded in the structure of the organization, not in formal training programs. Within this informal network of relationships, individuals learn a frame of reference that enables them to produce their work and to engage in activities in the managerial form sanctioned by the organization. The teaching process described above did not occur for the women in the B.C. study in the same manner that it did for men. Women at all organizational levels, from clerks to senior professionals, did not appear to participate in the organizational "team play" that was characteristic of the men. Men, in all job classifications, consistently referred to a sense of "we", describing their work as part of a team effort. They described their work in terms of the overall plan of their efforts, naming the divisional or branch objectives, other positions, and programs that formed the context of their work. The men rarely talked of their actual activities; rather, it was the direction in which their work would move them that was emphasized.  50 The responses of the women suggested that their work was carried out more in isolation. They readily described their daily activities, but rarely discussed where their work was heading or what it would do for them personally in terms of advancement. The authors of the B.C. Public Service report suggest that the differences between men and women noted above are the product of a social process and not a product of gender: This relation of work to personal advancement is learned in the ordinary course of work, through the activities of individuals and groups of individuals. It is also clear from the research that a great deal of training goes on in settings which have a more social character and in which women do not participate on the same terms as men. This might be, for instance, over a drink after work, or while travelling together on business. These are occasions on which being a woman is made relevant over and above being a colleague. Thus the occasions actually change as men accommodate women on this basis. This is often described as "women change the climate". Participation has a different character for women than for men. (British Columbia, MISBD, 1979, pp. 19-20) Accounts of explicit training about how to work in organizations were consistent across the stories of those who had advanced successfully at every level. The implications of this finding are that: People do not advance by themselves. They are advanced. This is largely the work of their superiors who have pushed and challenged them and taught them how to handle the organization. Teaching people what they need to know is an active enterprise. (British Columbia, MISBD, pp. 20-21) The authors conclude that if a great deal of this training takes place at an informal level, then "the inevitably social character of the conduct of the organization must be acknowledged and addressed as a fully proper area of organizational responsibility" (British Columbia, MISBD, 1979, p. 21).  4. Homosociality and Network Structure Several empirical investigations have documented the effects of homosociality on organizational network structure. The first of the studies reviewed here, while not conducted in an actual organization, does provide some useful evidence.  51 Larwood and Blackmore (1978) investigated the behaviour of 36 male and 24 female management students in soliciting volunteer leaders for a leadership task. They derived their predictions from the vertical dyad linkage model (Dansereau, Graen, & Haga, 1975) which conceptualizes a group as a series of dyadic connections between the leader and individual group members. The interactions within each dyad depend on the type of role relationship that is established between leader and member. If an "ingroup" relationship is formed, the group leader and group member make decisions together and the member assumes leadership functions with the blessings and support of the leader. If, on the other hand, an "outgroup" relationship develops, the leader will simply instruct the member how to behave. A close relationship will not develop between the two, the outgroup member will not be made part of the decision-making process, and is less likely to be designated a leader. As predicted, respondents were more likely to request same-gender than cross-gender acquaintances to volunteer for leadership tasks. This effect was stronger for men than for women. Also, as predicted, persons previously known to the respondent were more likely to be requested to take part in remunerative or challenging tasks. These results suggest that, given the continuing state of high-level positions in organizations being occupied primarily by men, more men are going to be solicited and sponsored for advancement within the management ranks than are women. Larwood and Blackmore (1978) also compared a postexperimental question with their behavioural data. The stated preference for asking a male rather than a female to participate was positively correlated with the actual number of male acquaintances requested to participate and negatively correlated with the number of previously unknown females requested to participate. The measure was unrelated to the number of previously unknown males requested to participate or the number of female acquaintances requested to participate. These findings would suggest that stated preferences for one gender over another (as measured by the Cross-Gender Preference and SameGender Preference subscales) may be correlated with reports of relationship behaviour (number of same- and cross-gender relationships reported, mentoring behaviours and relationship provisions supplied by these relationships).  52 Lincoln and Miller (1979) studied the effects of five personal attributes - authority, education, gender, race, and branch assignment -- on the instrumental (work contact) and primary (friendship) ties in five professional organizations. They attempted to account for the dyadic connections of persons within these networks of instrumental and primary ties by examining the attributes that each pair of individuals had in common. All employees of the five organizations were given questionnaires requesting them to list five persons in the organization with whom they had worked closely in the preceding month and five persons in the organization whom they considered to be their close friends. The response rate was 90% in four of the organizations and 8 5 % in the remaining organization. Lincoln and Miller (1979) chose as their unit of analysis the path distance or minimum number of links connecting each pair of individuals within each network. Using regression analysis, they examined the degree to which the attributes of authority, education, gender, race, and branch assignment predicted the proximity (path distance) of pairs of individuals. Their findings indicated that gender and race had a greater effect on friendship than on instrumental ties, with the tendency being to exclude women and nonwhites from the friendship networks. Overall, authority and education affected instrumental and friendship ties equally, serving to place persons highest in formal authority and education in central network positions. The attributes of authority, education, gender, and race tended to be status determinants: "White males with high education in formal positions of authority... [had] high probabilities of occupying the most central locations in the network space" (p. 193) and, hence, were most likely to receive both formal and informal communications about matters of import to the organization. Lincoln and Miller (1979) concluded by suggesting that friendship networks in organizations are not innocuous sets of social contact, rather: They are systems for making decisions, mobilizing resources, concealing or transmitting information, and performing other functions closely allied with work behavior and interaction [Since the attributes of gender and race] tend to exclude women and nonwhites from friendship networks, we might conclude that to the extent that friendship networks influence organizational process, reason gives way to prejudice, (pp. 196-197)  53 Rose (1985a) investigated the professional networks (departmental, university, and national) of assistant professors in psychology at 60 universities. The profession of psychology was chosen because, despite a continued higher proportion of men than women in the field, there were enough women (25%) from which to draw a sample. Forty-seven women and 43 men holding tenure-track positions completed questionnaires about the composition and functioning of their professional networks. Analysis of variance (Gender x Marital Status) was utilized to analyze network composition and effectiveness of network functions data. Rose (1985a) found fewer gender differences in junior faculty members' network composition and functioning than she had hypothesized. However, gender differences in network composition were found in three areas: women had more women colleagues than did men, women had a greater likelihood of having a higher-status woman associate than did men, and women had fewer associations with people in their previous institutions than did men.  Rose noted that the  women respondents in her study seemed to make up for their weaker ties to their graduate institutions by forming national networks equal in size to those of the men and by forming a small same-gender support network: "over 90% of the women had two or more same-sex associates, and close to two-thirds had a higher-status woman in their network" (Rose, 1985a, p. 544). Contrary to Rose's (1985a) hypotheses that women would have fewer higher-status associates and fewer collegial relationships with men, women listed approximately the same number of men in their network (M = 8.22) as did men (M = 8.26) and no gender differences were detected for "number of higher-ranking colleagues, number of higher-ranking male colleagues, or number of higher-ranking male colleague friends" (Rose, 1985a, p. 540). However, women did list as close friends significantly more higher-status women than men did. Marital status did not have a significant main effect. However, a significant Gender x Marital Status interaction effect was observed for mean proportion of colleagues listed as friends: "Single female faculty had the largest proportion of colleagues classified as close personal friends (M = .541), followed by married males, single males, and married females (M = .435, .402, and .378,  54 respectively)" (Rose, 1985a, p. 541). Rose's results suggest that single women faculty may integrate collegial and friendship relationships more than men and married women faculty do. Contrary to Rose's (1985a) hypotheses, no gender differences were found on items regarding the effectiveness of four (of five) network functions: friendship, socialization, career information, and access to current research information. However, gender differences were found on three measures assessing the effectiveness of respondents' networks in helping them to build a professional reputation (visibility function). Women rated their national network colleagues as significantly less effective than men's at increasing professional visibility. Women's colleagues were significantly less likely than men's to have recommended their work to others. Both women and men, though, did report receiving similar amounts of assistance from colleagues in terms of receiving career advice and being introduced to other colleagues. The final gender difference in the visibility function was that "compared to men, women rated their colleague-friends as having expended significantly less effort in helping them establish a network" (Rose, 1985a, p. 542). Rose noted that because women did not differ from men in terms of expectations for networks to promote visibility, the above-noted differences in perceived effectiveness of network functioning cannot be accounted for by women having higher expectations than men.  No gender differences were found  on measures of career success either, suggesting that lower performance by women was not a reason for differential treatment by colleagues. Rose's (1985a) findings contradict, in some ways, previous findings of women's exclusion from men's networks. It could be that both the men and women psychologists who responded to her questionnaire had become more aware of the importance of "networking" than people in previous studies had been. Also, given that the percentage of women in psychology is 25%, issues of isolation and discrimination related to being a token woman (Kanter, 1977) may have been reduced. It should be noted that the women in Rose's study did have more women in their networks than did men, indicating the establishment of a small same-gender support network by the majority of women.  55 Brass (1985) mapped the formal and informal networks of men and women in one organization and studied the relationship of these interaction patterns to promotions and perceptions of organizational influence. One hundred forty full-time, nonsupervisory employees (76 men, 64 women) of a newspaper publishing company completed a questionnaire about their organizational network involvement. To obtain information about the formal or workflow networks, respondents were asked to name all persons who supplied them with work assignments and to whom they distributed completed work. Information about the informal interaction networks was obtained by respondents listing persons to whom they talked often about work-related matters and whom they considered to be close friends. Persons named in this manner were counted as network members only if they reciprocated the original respondent's choice. Workflow networks were also verified by direct observation and by interviews. Many measures of workflow and interaction network centrality were examined: centrality within the entire organization, within the department in which the respondent worked, and within subunits of employees with the same immediate supervisor. In addition, centrality within men's and women's networks was assessed by using all men in the organization as the reference group for the men's interaction network centrality scores and all women in the organization as the reference group for the women's interaction network centrality scores. Finally, access to the dominant coalition -- a small group of the most influential top-level personnel -- was assessed. Following Blau and Alba (1982), Brass (1985) defined centrality as: the minimum distance between a focal person and all other persons in the pertinent reference group. The distance was measured by counting the number of links between a focal person and each other person; the resulting total was the centrality score, which was then divided by n-1, where n equalled the number of persons in the reference group, (p. 332)  Given this definition, Brass suggests that a person with a high centrality score would have easy access to others in his or her reference group. The findings indicated that, in this organization comprised of approximately 4 0 % women and 6 0 % men, women were more central than men in both the workflow and interaction networks  56 when the whole organization was the reference group. However, "there appeared to be two informal, [gender-]segregated networks operating in the organization" (p. 339). When naming members of their informal interaction network, "men listed other men 75 percent of the time, and women listed other women 68 percent of the time" (p. 336). Women were not as central as men in the all-male interaction network and men were even less central in the women's interaction network. In addition, women were less central than men in informal interactions with the dominant coalition composed of four top-level men. Amount of perceived influence was assessed by peer and supervisor ratings. The overall pattern of results indicated that perceived influence was strongly related to integration into social networks. Influence was significantly related to integration into the dominant coalition; this relationship was particularly strong for women. For both women and men, centrality in cross-gender interaction networks was significantly related to supervisors' ratings of influence while centrality in same-gender networks was not. Thus, for both women and men, contacts with cross-gender persons were advantageous in terms of perceived influence. In terms of influence, it was particularly important for women to be part of integrated (mixedgender) workgroups. Women in integrated workgroups scored significantly higher on several measures than did women in all-female workgroups. These measures were: centrality in subunit and departmental interaction networks, supervisors' ratings of influence, contacts with persons outside the immediate workgroup, centrality in men's interaction network, and access to the dominant coalition. When compared with their male counterparts in mixed-gender workgroups, these women were not significantly different from the men; the exceptions were their scores on three measures. These women scored significantly higher than the men on centrality in departmental, subunit, and women's interaction networks. While the overall pattern of Brass' (1985) results indicated gender segregation in the informal interaction networks, women in mixed-gender (integrated) workgroups were the exception to this pattern, as noted above. Men in the integrated workgroups were also more central to the women's interaction network than were men in all-male workgroups. Thus, proximity or the presence of  57 cross-gender persons in the immediate workgroup was an important factor in being integrated into cross-gender interaction networks. However, proximity alone could not explain differences in informal interaction network centrality. For example, "the men in these integrated workgroups were still significantly less central to the women's network than women who were also members of integrated workgroups" (p. 340). The investigator suggests that it may have been more important (in terms of influence and career advancement) for the women in mixed-gender workgroups to interact with men than it was for the men to interact with the women. Multiple regression analysis indicated that "for women, access to the dominant coalition and centrality in the men's interaction network were significantly related to both influence measures [supervisors' ratings and nonsupervisors' listings] when all other variables were controlled" (p. 338). In the three years following the initial data collection, seven men and three women from the original sample were promoted. Correlations between promotions and centrality for the entire sample indicated that promotions were significantly related to centrality within the men's, departmental, and dominant coalition informal interaction networks. In terms of attitudinal homosociality, it may have been that the women in these mixedgender workgroups had a higher level of Cross-Gender Preference than did the men. Note that Cross-Gender Preference refers to a preference for cross-gender persons in situations involving cross-gender persons of higher status and power. A woman seeking to increase her own sphere of influence and to maximize her chances of promotion might seek relationships with those perceived as having the power to assist her in those pursuits. In the case of the women in Brass' (1985) study, it was access to the men's and dominant (male) coalition interaction networks that made the difference in terms of influence and promotions. The women in mixed-gender workgroups also may have maintained moderate to high levels of Same-Gender Preference (preference for same-gender persons in situations of equal status) and Valuing (valuing of same-gender relationships in and of themselves). Hence, they remained more central than men in the women's interaction networks, perhaps having different needs met in their relationships with other women.  58 In Brass' (1985) study, fewer women than men received promotions and women were perceived by supervisors and nonsupervisors as being less influential than men. Men seeking increased influence and promotions, then, might have indicated higher levels of Same-Gender Preference and Valuing and lower levels of Cross-Gender Preference. In general, cross-gender persons (women) perhaps were perceived as unnecessary or unhelpful in terms of increasing influence and career advancement. It should be noted, though, that in Brass' study, men who were more central to the women's interaction network were perceived as having more influence than were other men. Given that it was the men in integrated work groups who were more central to the women's interaction network and that the women in these integrated workgroups were perceived as equally influential to men, it may be the specific women to whom the men related that caused an increase in perceived influence. Brass (1985) contends that "it is impossible to ascertain whether men actively exclude women from their informal interactions, and vice versa, or whether members of each gender exclude themselves from ties with members of the other gender" (p. 340). The current investigator suggests that it is not entirely impossible to ascertain the above. Use of the Bonds Between Women/Bonds Between Men Scale to measure aspects of homosocial attitudes (Cross-Gender Preference, SameGender Preference, Valuing, Identification, and Jealousy) could contribute information about factors hypothesized to be related to inclusion in or exclusion from same- and cross-gender networks. Brass (1985) also suggested that "further research should deal with the specific content of [network] interactions" (p. 341). Measures of mentoring and friendship could be utilized to examine in detail the content of network interactions. The theory and research presented in this section have pointed clearly to the importance of supportive relationships in the workplace. The skills of relating successfully to the work of the organization and, hence, advancing one's career within the organization, are taught within the context of supportive relationships that are largely informal in nature. Persons lacking these relationships - for example, the women described by Abramson (1979) and the British Columbia, MISBD (1979) report -- had vastly different outcomes than did the women and men who had  59  supportive relationships. Those persons lacking supportive relationships experienced frustration, disappointment, and blocked advancement, while those persons who had supportive relationships successfully advanced to senior levels within the organization. The body of literature presented in the next section expands upon the concept of the importance of supportive relationships at work. The topic is mentoring.  D. MENTORING The actions engaged in by the male bosses of the 25 successful women in Hennig and Jardim's (1977) study have more recently come to be known as mentoring. Kram (1985) describes the mentor relationship as: The prototype of a relationship that enhances career development.... [It is] a relationship between a young adult and an older, more experienced adult that helps the younger individual learn to navigate in the adult world and the world of work. A mentor supports, guides, and counsels the young adult as he or she accomplishes this important task. (p. 2)  It should be noted that the mentoring relationship is the prototype of supportive work relationships. Understanding this relationship can provide valuable information about the effect of work relationships on career development (Kram, 1985). However, mentoring behaviours also may be received from persons other than a full-fledged mentor. For example, supportive work relationships with peers may be an important alternative to a relationship with a mentor (Kram, 1985; Kram & Isabella, 1985). More will be said about this later following the discussion of the mentoring relationship per se. Many writers and researchers (e.g., Alleman, 1985, 1986; Farren, Gray, & Kaye, 1984; Henderson, 1985; Kram, 1985; Missirian, 1980/1981; Reich, 1985,1986; Schroeder, 1988/1989; Zey, 1984) have described, in greater or lesser detail, the diverse roles, functions, and behaviours engaged in by mentors in support of their proteges. Kram (1985) developed a classification scheme of mentoring behaviours using information gathered during in-depth interviews about the nature of relationships between junior and senior managers in an organizational setting. Until the recent work  60 of Schroeder (1988/1989), Kram's was the most comprehensive listing and description of mentoring behaviours to date. Using Kram's (1985) typology as an organizing theoretical framework, and drawing from the work of others who had delineated mentoring functions, Schroeder (1988/1989) developed a measure of supportive mentoring behaviours which he used in a study of the effectiveness of mentoring behaviours. The following discussion of mentoring roles, functions, and behaviours draws heavily from the work of Kram (1985) and Schroeder (1988/1989) because they are the most comprehensive models presented to date.  1. Mentoring Functions Kram (1985) divided mentoring functions into two general classifications - career functions and psychosocial functions. Career functions are defined as "those aspects of the relationship that enhance learning the ropes and preparing for advancement in an organization" (Kram, 1985, p. 22). Psychosocial functions are defined as "those aspects of the relationship that enhance a sense of competence, clarity of identity, and effectiveness in a professional role" (Kram, 1985, p. 22). The career functions assist advancement within the organizational hierarchy, while the psychosocial functions affect the individual on a more personal level. Provision of career functions require that the senior person have enough experience and influence within the organization to provide the following individual functions: sponsorship, exposure and visibility, coaching, protection, and challenging assignments. Psychosocial functions, in contrast to career functions, require a closer interpersonal relationship based on mutual trust and increasing levels of intimacy. When the psychosocial functions of role modeling, acceptance and confirmation, counselling, and friendship are present:  Each individual experiences acceptance and confirmation through interaction with the other; mutual liking and respect support the young adult's views of self in the new work role, and simultaneously support the senior adult's views of self as someone with valuable wisdom and experience to share. (Kram, 1985, p. 23) According to Kram, a given relationship need not provide both career and psychosocial functions. Relationships that provide only career functions are less intimate and have a weaker interpersonal bond than do relationships that include psychosocial functions. Relationships that provide only  61 career functions are valued mainly for the instrumental needs that they meet, while relationships that include psychosocial functions are viewed as more unique and indispensable than most other work relationships.  Each of the career and psychosocial functions is described individually below.  Career Functions The career functions presented below are those adapted by Schroeder (1988/1989). They are very similar to the functions discussed by Kram (1985): Sponsoring, Exposure and Visibility, and Protection are the same, Kram's coaching is included under Schroeder's Teaching the Informal System, and Kram's challenging assignments has been subsumed under Schroeder's Teaching the Job. Sponsoring. Schroeder (1988/1989) defined Sponsoring as "communicating information to others, either verbally or nonverbally, concerning the potential or competence of an employee" (p. 5), thereby ensuring that employees will be considered for promotions and career opportunities.  Exposure and Visibility. This function differs from Sponsoring in that with Exposure and Visibility, the protege is involved in the actions taken. That is, the senior person provides the employee with "tasks or opportunities that ensure that others will become personally acquainted with that employee" (Schroeder, 1988/1989, p. 6). It is then up to the employee to create his or her own impression by engaging in the task.  Teaching the Job.  The senior person provides the employee with "knowledge or  experiences that help an employee learn technical and managerial skills" (Schroeder, 1988/1989, p. 7).  62  Teaching the Informal System. This function helps employees to gain important knowledge about how to work effectively within the organizational structure (Kram, 1985). As defined by Schroeder (1988/1989), it involves the "imparting of knowledge to an employee concerning informal aspects of the organization such as norms, mores, and politics" (p. 8).  Protection. As defined by Schroeder (1988/1989), Protection has two aspects. First, a provider of this function may prevent "an employee from being exposed to tasks, positions, or situations in which the employee is unlikely to succeed" (p. 9). Second, the senior person may protect "an employee who has engaged in behaviours or actions that may be detrimental to the employee by taking the blame for or not telling others about these actions or behaviours" (p. 10).  All of the above definitions of career functions were adopted for use in the current study.  Psychosocial Functions Schroeder (1988/1989) chose a much narrower definition of psychosocial functions than was adopted in the current investigation. Because it was hypothesized that significant gender differences in supportive work relationships would be noted in the psychosocial domain, measurement of the psychosocial aspect of relationships was expanded in the present study. However, because Schroeder's (1988/1989) measure was the measure of mentoring behaviours used in the present study, the definitions of the psychosocial functions adopted by him are presented below. Following that, Kram's (1985) description of the psychosocial functions is outlined as a starting point for the presentation of the expanded view of psychosocial functions adopted in the current study. The psychosocial functions as defined and measured by Schroeder (1988/1989) are:  Role Modeling. By acting as role models, senior persons provide employees with opportunities to learn by observation.  63  Encouragement. Schroeder (1988/1989) defined Encouragement as "providing encouragement or support that helps an employee feel competent, confident, or worthwhile" (p. 14).  Personal Counselling. This function is defined as "discussing an employee's concerns involving self, career, or family (Schroeder, 1988/1989, p. 15).  It should be noted that each of the psychosocial functions as defined by Schroeder involves behaviours that the senior person engages in on behalf of the employee. No mention is made of a reciprocal relationship. Kram's (1985) description of the psychosocial functions and the expanded definitions adopted in the present study both include considerable reciprocity in the relationship. Kram's (1985) definitions and descriptions of the psychosocial functions are as follows:  Role Modeling. Kram (1985) states that role modeling is the psychosocial function which is reported most often. The senior person's values, attitudes, and behaviours provide a model for the junior person to imitate. The junior person's basic admiration and respect for the senior person makes the younger person open to learning by observation. It is the emotional attachment that is formed between the two that makes role modeling effective. The level of emotional attachment and identification may cause difficulties in the relationship, though, as both persons may feel "attachment, protection, ambivalence, and rebellion that parallel experiences in earlier [family] relationships" (Kram, 1985, p. 34). The process of identification is a complex one. The junior person may identify with and emulate some aspects of the senior person's style and reject other aspects. Ideally, and over time, the younger person will differentiate him or herself from the admired senior person by incorporating valued aspects of the person into his or her own self structure and by choosing to remain different in other aspects.  64 Acceptance and Confirmation. Kram (1985) contends that, through this function, both persons receive support and encouragement; both "derive a sense of self from the positive regard conveyed by the other  Positive feedback on performance, mutual liking, and mutual respect  help both individuals" (p. 35). The acceptance and confirmation provided by the relationship enables the junior person to take risks and try out new behaviours in his or her work world. This function enables the expression of differences and self-differentiation.  Counselling. Through this function, the junior person is enabled to discuss personal concerns that might detract from effective work and a positive sense of self as a member of the organization. Personal concerns will vary depending on the career stage of the individual. For example, a person in the early stage of a career might want to discuss issues of competence and potential or how work commitments can be balanced with other aspects of life. Other concerns may surface in later years. Kram (1985) states that: What is shared in the relationship goes beyond the boundaries of most hierarchical relationships. The young adult derives comfort in discovering that he can share doubts and concerns without risking exposure to others in the organization. He can also address conflicts that would otherwise interfere with effectiveness and selfworth. The senior adult satisfies important needs by enabling a younger colleague to successfully manage personal dilemmas. And by sharing his own experiences and self-insights, he remembers previous points of decision during earlier career stages. The alliance formed through the counseling function counteracts the organizational force that can contribute to alienation and a decline in self-worth, (pp. 37-38) Friendship. According to Kram (1985), "this function is characterized by social interaction that results in mutual liking and understanding and enjoyable informal exchanges about work and outside work experiences" (p. 38). As this function emerges, increasing mutuality is experienced in the relationship. Kram suggests that the friendship function is limited in relationships between junior and senior members of an organization. Conflicts may arise if the roles of boss and friend are enacted simultaneously. Therefore, most of the people she interviewed chose to limit social interaction to the work setting while continuing to enjoy sharing personal experiences at work, eating lunch together, and so forth.  65 There is considerable overlap in the four psychosocial functions described by Kram (1985). The common theme among them is the presence of emotional bonding, a theme that is not present when the career functions are present alone. Schroeder's (1988/1989) instrument measures a part of each of the psychosocial functions of role modeling, acceptance and confirmation (he calls this Encouragement), and counselling (he calls this Personal Counselling). As stated previously, though, Schroeder's (1988/1989) measure does not tap into the reciprocal nature of emotional bonding that may occur when the psychosocial functions are present. Alleman's (1985) measure of mentoring behaviours (The Alleman Leadership Development Questionnaire) also contains scales that tap the psychosocial domain. She has called these scales General Counselling, Demonstrated Trust, Associate Socially, and Friendship. While many of the items in her scales address important aspects of the more personal side of the mentoring relationship, many of the items are too situation-specific to apply to a wide range of people. For example, two items in Associate Socially concern helping the other person with a political campaign and voting against the other person in an election. Another problem with the Leadership Development Questionnaire is its psychometric properties. Analyses reported in the questionnaire manual (Alleman, 1985) were based on data gathered from only 100 respondents and factor analysis had not been done to validate the structure and content of the scales. For the purpose of his study, Schroeder (1988/1989) chose not to include friendship as a mentoring function. He based this decision on the views of writers such as Klopf and Harrison (1981), Missirian (1980/1981), and Reich (1986) who suggest that friendship is a potential outcome rather than function of a mentor relationship. As discussed earlier, Kram (1985) noted the difficulty of friendships between superiors and subordinates and commented that the friendship function was limited in mentoring relationships. Her findings indicated that friendship was most likely to come into being in the redefinition phase of a four-phase mentoring relationship (initiation, cultivation, separation, redefinition). While Schroeder (1988/1989) did not include friendship as a mentoring function, he did measure (with a 5-item scale) the level of friendship present in the supportive relationships he studied. He found a significant positive correlation between the level of supportive  66 behaviours received from a person and the level of friendship reported with that person, indicating that friendship is, indeed, an aspect of supportive work relationships. Analysis of variance was used to compare the level of friendship received from supervisors, persons at a higher level than supervisors, and co-workers. The results indicated that "the level of friendship with co-workers was significantly higher than the level of friendship with supervisors or with people at hierarchical levels higher than supervisors" (Schroeder, 1988/1989, p. 140). This finding makes sense given that opportunities for a reciprocal relationship are likely to be more frequent among persons of equal rank. Friendship was present in the relationships, though. It was just more limited in supervisorsubordinate relationships. Schroeder's (1988/1989) findings in regard to the positive relationship between friendship and mentoring behaviours, and the findings of others such as Kram (1985) and Missirian (1980/1981) in regard to the presence, advantages, and disadvantages of deep emotional bonding in some mentoring relationships, suggests that the psychosocial dimension is worthy of further study. While two people involved in a mentoring relationship may choose to limit the amount of social interaction they have outside the immediate work setting, particularly during the time when the senior person has supervisory responsibilities with the junior person, the two are often emotionally bonded to each other. In her study of the mentoring process in the career development of women managers, Missirian (1980/1981) found that "during all three stages of the mentoring process, a richness of rapport, a caring, a trust develops between mentor and protege which can be described as a love relationship" (p. 115). As in any love relationship, the relationship is not free of conflict. The degree of emotional bonding present and the depth at which each affects the other means that both will, at times, be vulnerable to the other, particularly during periods of tension in the relationship. Missirian (1980/1981) suggests that "it is the ability of mentor and protege to confront one another on issues and yet maintain their respect and affection for one another that test the real strength of the relationship and the participants" (p. 115). Reich (1986) also has commented on the difficulties experienced in some mentoring relationships. While half of the men and women he studied listed few or no negative aspects of  67  mentoring relationships, others did cite major drawbacks. The main disadvantage listed by both women and men was being so closely identified with the mentor that others perceived the protege as the mentor's "person". Other drawbacks listed were being protected too much and being kept from better positions. Sixteen percent of the women and 2 2 % of the men stated that the mentoring relationship had caused "a moderate amount of stress and that one or more of the ... [drawbacks listed above had] truly hurt the relationship" (Reich, 1986, p. 53). For example, the close identification and labelling discussed above was harmful in two ways. First, it limited opportunities for the protege to work with other people. Second, in some cases, attempts by the protege to achieve what one woman in Reich's (1986) study called "independence versus imitation" (p. 53) harmed both the mentoring relationship and the protege's career. Mentoring relationships that are rich in the psychosocial domain present both a risk and an invaluable opportunity for personal and professional growth for both the mentor and the protege (Missirian, 1980/1981). A greater understanding of the psychosocial dimension of mentoring relationships could help to minimize the risks and maximize the chances for successful outcomes of mentoring activities. Prior to the current study, a comprehensive and psychometrically strong measure of the psychosocial dimension of mentoring relationships had not been employed in a study of supportive work relationships. The career functions and psychosocial functions, while conceptually distinct, have not yet been found to be empirically distinct. As described in Chapter III (Method) of the current document, the results of principal component analyses of the eight subscales (individual mentoring functions) in Schroeder's (1988/1989) measure indicated one general factor. Relatively high interscale correlations supported the conclusion that the mentoring functions, as defined and operationalized by Schroeder, formed one general factor. The present investigator suggests that one reason why career and psychosocial functions have not yet been found to be empirically distinct is because the definition and subsequent measurement of the psychosocial functions has been too narrow. Ragins and McFarlin's (1989) mentor role instrument (MRI), published after the time of data collection for the current study, did provide a more thorough coverage of psychosocial provisions as defined by Kram (1985). The instrument measured the  68 following psychosocial provisions or roles (three items per role): friendship, role model, counselling, acceptance, social, and parent. However, the instrument was developed using a pool of only 69 proteges. The present study added to the body of research on mentoring and other supportive work relationships by employing the Close Social Relationships Scale (CSRS) (Woolsey et al., 1988b) to measure, in an expanded way, the psychosocial dimension of these relationships. The CSRS measures both conceptually and empirically distinct aspects of the psychosocial provisions discussed by Kram (1985). For the purpose of the present study, these dimensions were called relationship provisions. The relationship provisions measured by the version of the CSRS used in the present study are:  Intimacy.  When this provision is present, the two persons are able to share with each other  their deepest, most private feelings; they exchange confidences and openly express their affection for one another. Intimacy is discussed by Kram (1985) under counselling and acceptance and confirmation.  Similarity.  The two persons are drawn together because they perceive many important  similarities between them. Similarities may be noted, for example, in values, interests, and personality. Kram's (1985) description of the identification process which occurs in role modeling seems to be very close to the Similarity provision.  Defiance of Convention.  When this provision is present, the persons feel so comfortable  with each other that they are able to defy conventional standards of appropriate behaviour when they are together (e.g., they are able to be "wild and crazy" when together). This dimension of close social relationships emerged from an interview study about close same-gender relationships (Woolsey & McBain, 1987b). In a subsequent factor-analytic study, Defiance of Convention was validated as a separate dimension (Woolsey et al., 1988b). To date, the investigator has not located  69 any other literature that discusses this dimension. Given that supervisors or senior persons in the organization may represent authority and conventional standards, it seems reasonable to assume that this dimension may be more present in close peer relationships than in supervisor-subordinate relationships.  Emotional Support. The two persons provide each other with a sense of comfort, encouragement, and emotional support. They admire each other and learn things from each other. This dimension is most comparable to Kram's (1985) acceptance and confirmation function.  Respect for Differences. When this provision is present, the two persons allow each other to be separate and different individuals -- to excel at different things, to grow and change, and to make their own decisions. Kram (1985) discusses this provision under two psychosocial functions -role modeling and acceptance and confirmation. Under role modeling, the expression of differences occurs when the junior person differentiates him or herself from the admired senior person. In doing so, the junior person incorporates into his or her concept of self those aspects of the senior person that are valued, and makes a decision to remain different in regard to the less desired aspects. Kram (1985) also suggests that it is the presence of the function of acceptance and confirmation that enables the expression of differences and self-differentiation.  Lifetime Attachment. When this provision is present, the two persons state that they have developed a unique and special relationship with each other that has stood the test of time. Kram (1985) does not discuss this provision as a separate psychosocial function. Rather, she states that relationships that provide psychosocial functions in addition to career functions are "characterized by greater intimacy and strength of interpersonal bond and are viewed as more indispensable, more critical to development, and more unique than other relationships in the manager's life at work" (Kram, 1985, p. 24).  70 2. Gender Differences in Mentoring Relationships The results of some investigations have suggested that women managers may require and receive different types of mentoring than do men (Ragins, 1989). For example, several studies have found evidence of gender differences in mentoring in regard to the psychosocial domain. Collins' (1983) study of professional women and their mentors led her to conclude that women seem to need and appear to be given more support, encouragement, and instilling of confidence, while men seem to need and appear to receive training in risk-taking, leadership skills, and giving direction. Burke's (1984a) findings indicated that "there appeared to be unique features to the mentoring relationship (i.e., psychosocial functions) when females were involved either as mentors or proteges" (p. 368). Burke concluded that mentoring relationships involving women appeared to be comparable to those involving men; however, women's relationships included personal and emotional aspects as well. Reich (1986) also found that women's mentor-protege relationships were different from relationships involving men only, in that "the affective, or emotional, quality was more vital for women than for men" (Reich, 1986, p. 53). The present study was a more complete test of gender differences in the psychosocial aspect of mentoring relationships than were the three studies cited above. Reich (1986) compared the mentoring experiences of female and male executives. He found that 7 7 % of the women and 9 0 % of the men had had a mentor who had influenced their career development. Seventy-one percent of the women had had male mentors. Most of the women studied had had mentors by the age of 37 (compared to age 35 for the men) and had, at that time, begun to serve as mentors to others. Only 50% of the women were their proteges' supervisors, in contrast to 7 4 % of the men. Reich's (1986) most significant finding was that 90% of the women who had become mentors were mentors to other women; fewer than 5 % of the men had women proteges. These findings are consistent with Larwood and Blackmore's (1978) finding that both women and men prefer same-gender persons to cross-gender persons when selecting others for leadership tasks. Henderson's (1985) results indicate a similar pattern. He found that "women executives were three times more likely than a man to have a woman as a mentor" (p. 860). The investigator suggests that one reason for these findings may be the pattern of homosocial attitudes  71 held by the women and men studied. For example, the women in Reich's (1986) study may have had high levels of Same-Gender Preference, Valuing, and Identification and low levels of Jealousy. Reich (1986) did not report how many of the women served as mentors to men, so it is difficult to hypothesize what the women's Cross-Gender Preference scores may have been. The men in his study might have had low levels of Cross-Gender Preference and high levels of Same-Gender Preference and Valuing. In his review paper on women and mentoring, Noe (1988) stated that "the number of mentoring relationships (mentorships) available to women does not appear to be keeping pace with the growing number of women needing mentors" (p. 65). Gender differences in relationship styles such as those noted by Collins (1983), Burke (1984a), and Reich (1986), and the preference of both men and women for relationships with same-gender persons may contribute to this deficit in mentoring relationships for women. Given that higher level positions in most organizations continue to be occupied mainly by men, women may lose out on valuable mentoring opportunities because of homosocial preferences and behaviours.  3. Outcomes of Mentoring Behaviours A growing body of literature and research attests to the importance to career development of establishing a relationship with a mentor (Collins & Scott, 1978; Fagan & Fagan, 1983; Gray & Gray, 1986; Hunt & Michael, 1983; Klopf & Harrison, 1981; Kram, 1985; Levinson, Darrow, Klein, Levinson, & McKee, 1978; Noe, 1988; Orth, Wilkinson, & Benfari, 1987; Reich, 1985,1986; Roche, 1979; Zey, 1984). To date, however, "few rigorous quantitative studies of the ... outcomes of mentoring have appeared in the organizational behavior literature" (Noe, 1988, p. 66). The current study and the recent work of Schroeder (1988/1989) have addressed this deficit. Schroeder (1988/1989) investigated the effectiveness of mentoring behaviours by examining the relationship between employee outcomes and the level of mentoring behaviours received. He collected data for 11 individual career- and job-related outcome measures: Organizational Commitment, Role Ambiguity, Role Conflict, Job Satisfaction, Acceptance by Co-  72 workers, Job Skill Development, Interpersonal Skill Development, Conceptual Skill Development, Satisfaction With Progression, Average Number of Promotions per Year, and Average Annual Salary Increase. The results of principal component analysis performed on the correlation matrix for the 11 individual outcome measures indicated three separate components. The first principal component was labelled the Job-Related Composite. It consisted of Organizational Commitment, Role Ambiguity, Role Conflict, Job Satisfaction, and Acceptance by Co-workers (all of these measures loaded highest, and above .40, on the Job-Related Composite). The second component, labelled the Skill Development Composite, had high loadings for the three skill development outcomes (Job, Interpersonal, and Conceptual Skill Development). The third component - the Promotional Composite - had high loadings for Satisfaction With Progression, Average Number of Promotions Per Year, and Average Annual Salary Increase. The first principal component for the measures associated with each composite was extracted and component scores were computed using the regression method. Scores for each of the composites were standardized to a mean of zero and a standard deviation of one. Internal consistencies (alpha) for the composites were .77 for the JobRelated Composite, .58 for the Skill Development Composite, and .57 for the Promotional Composite. Schroeder (1988/1989) found significant positive correlations between each of the outcome composites and the level of mentoring behaviours received. The highest correlations were between the Skill Development Composite and the level of supportive behaviours received (e.g., for the first source of supportive behaviours for the entire sample, r = .47, p < .001, n = 372). These findings indicate that supportive mentoring behaviours do have a positive relationship with career development. The current study employed many of the outcome measures used by Schroeder (1988/1989). The skill development outcomes (job, interpersonal, and conceptual skills) were chosen for use in this study because of the strong relationship between each of the individual scales and the Total Supportive Behaviour Score (TSBS) obtained on Schroeder's measure of supportive (mentoring) behaviours, and between the composite score (Skill Development Composite) and the  73 TSBS. Reich (1985,1986) also has noted the relationship between the receipt of mentoring behaviours and protege skill development. All of the outcomes comprising the Promotional Composite (Satisfaction With Progression, Average Number of Promotions per Year, Average Annual Salary Increase) also were adopted for use in this study. While the relationships between the Promotional Composite and mentoring behaviours were not as strong in Schroeder's (1988/1989) study as those between Skill Development and mentoring, they were still significant relationships (e.g., for the first source of supportive behaviours for the entire sample, r = .28, p < .001, n = 355). Other studies also have noted the relationship between inclusion in informal networks and/or receiving mentoring behaviours and promotions (Brass, 1985; British Columbia, MISBD, 1979; Hennig & Jardim, 1977; Reich, 1986). Only some of the outcomes which composed Schroeder's Job-Related Composite were included in the present study. Organizational Commitment was chosen because of the positive and significant correlation between it and mentoring behaviours (e.g., for the first source of mentoring behaviours, r = .39, p < .001, one-tailed). Mowday, Steers, and Porter (1979) contend that commitment "involves an active relationship with the organization such that individuals are willing to give something of themselves in order to contribute to the organization's well being" (p. 226). Individuals who are not actively involved in giving and/or receiving support to/from other organizational members are unlikely to feel a high degree of organizational commitment. Schroeder's (1988/1989) Acceptance by Co-workers also was used in the present study. Peer acceptance has been cited as a concern during the establishment phase of career development (Gould & Hawkins, 1978; Mount, 1984; Stumpf & Rabinowitz, 1981). Acceptance by Co-workers can be conceptualized as one measure of the degree to which an individual has been integrated into informal networks. Job Satisfaction also was employed as an outcome variable in the present study, although a different (shorter) measure than that employed by Schroeder (1988/1989) was used. Individuals are likely to feel more general satisfaction with their jobs if they have become part of a network of relationships in their place of employment. Finally, an outcome variable (Loneliness) not  74 considered by Schroeder (1988/1989) was employed. Research has shown that loneliness "reflects an individual's subjective perception of deficiencies in his or her network of social relationships. These deficiencies may be quantitative (e.g., not enough friends) or they may be qualitative (e.g., lacking intimacy with others)" (Russell, Cutrona, Rose, & Yurko, 1984, p. 1313). The present study examined the supportive behaviours received from social networks in the workplace. The degree of loneliness experienced by respondents was utilized as one measure of the adequacy of these networks in meeting individual relational needs. Schroeder's (1988/1989) conceptualization of the three-part structure of employee outcomes was used to group conceptually the 10 outcome variables used in the present study. Given that some variables were deleted from and added to Schroeder's list, the factor structure of the outcome variables investigated in this study needed to be confirmed by factor analysis. The results of these analyses are presented in Appendix C (Preliminary Analyses).  4. A Network of Supportive Relationships as an Alternative to a Single Mentor Mentoring behaviours need not be obtained only from someone who is perceived to be a mentor. For example, 35 of the 397 respondents in Schroeder's (1988/1989) study received relatively high levels of supportive (mentoring) behaviours from persons they did not view as mentors according to two definitions of mentoring often cited in the mentoring literature. The mentoring relationship may be seen as a prototype of supportive work relationships (Kram, 1985). Supportive behaviours also may be received from sources other than mentors or superiors. In Schroeder's (1988/1989) study, supportive behaviours were received both from co-workers and from persons above the respondent in organizational rank. Kram (1985) contends that it is a misconception to believe that "finding a mentor is the key to individual growth and career advancement" (p. 199). She suggests that:  Given the limitations of mentor relationships, the fact that they are relatively unavailable to most individuals in organizations, and their potential destructiveness in certain situations, it is risky to rely on one individual for all developmental functions. Relationships with peers can also offer developmental functions, and  75 individuals should develop a relationship constellation that consists of several relationships, each of which provides some career and/or psychosocial functions. (Kram, 1985, p. 200) Kram suggests that one's relationship constellation can include friends, family members, peers, bosses, and mentors. Other writers and researchers have expressed similar sentiments. The data of Harlan and Weiss (1980) revealed that: Very few men and women have mentoring relationships. What they described as mentoring was, in fact, multiple helping relationships that facilitated their movement upward having one particular person who is involved closely and intently with another person's career is unlikely to exist. However, a broader network of helping relationships does appear to occur for access to job opportunity information for both men and women, (pp. 194-195) Darling (1985), too, found that full-blown mentors were rare. A much more frequent occurrence was "a succession of less intense, less encompassing relationships that, in conjunction with mentoring events, add up to very formative mentoring experiences" (Darling, 1985, p. 41). Shapiro, Haseltine, and Rowe (1978) suggest that mentors are at one end of a continuum of supportive/advisory relationships composed of peer pals, guides, sponsors, and mentors. All of these relationships facilitate job entry and career mobility. Missirian (1980/1981) conceptualized a similar continuum based on the degree of power and access to resources possessed by the support persons. She called these persons peers, coaches, sponsors, and mentors. The Woodlands Group (1980) discussed the importance of three of these roles (coach, sponsor, and mentor) in the development of managers. Phillips-Jones (1982) distinguished between primary mentors, who "significantly help you reach your major life goals  [and leave] an indelible mark on your career" (p. 21), and secondary  mentors who exert less influence on one's career but are helpful with certain specific needs. Secondary mentors include peer strategizers (co-workers, peers, friends, neighbours, lovers, and other acquaintances who assist in planning and implementing career goals), unsuspecting-hero role  76 models (heroes from public life, books, and so forth), and career favour-doers (persons who provide career favours, often on a one-time basis). Given that full-blown mentors are uncommon (Darling, 1985; Harlan & Weiss, 1980; Kram, 1985) and that mentoring functions which benefit the recipient can be received from various other individuals in the workplace, the present study examined the level of mentoring behaviours received from many different sources. The receipt of these behaviours was measured by Schroeder's (1988/1989) supportive (mentoring) behaviours measure. The behaviours measured by this scale are referred to as mentoring behaviours in this study. Respondents were requested to list persons in their organization who provided each of the eight mentoring functions outlined earlier in this section (see Mentoring Functions subsection). Then, because the level of mentoring behaviours was hypothesized to differ according to the gender of the recipient and the gender of the persons providing the supportive behaviours, respondents were requested to complete Schroeder's (1988/1989) scale twice: once with reference to the men listed, and once with reference to the women listed. Additional information about the psychosocial functions provided by these men and women was obtained by respondents completing the Close Social Relationships Scale (Woolsey et al., 1988b) in regard to those persons in their networks with whom they had more personal relationships. In this study the functions measured by the Close Social Relationships Scale are referred to as relationship provisions.  E. THE STAGES OF CAREER DEVELOPMENT Mentoring and friendship needs vary according to the life and career stage of the individual (Kram, 1985). Therefore, theory and research on career stages are reviewed briefly in this section. Hall's (1976) composite model of career stages, based on the work of Erikson (1963), Super (1957), Hall and Nougaim (1968) and others, divides the adult career into four stages according to age: establishment (up to age 30), advancement (ages 31 to 44), maintenance (ages 45 to 64), and decline (age 65 and beyond). This model has been supported by recent research on organizational  77 careers using an all-male sample of professional accountants (Adler & Aranya, 1984) and a sample of salespersons not identified according to gender (Slocum & Cron, 1985). Other researchers (Gould & Hawkins, 1978; Mount, 1984; Stumpf & Rabinowitz, 1981), drawing primarily from the work of Hall and Nougaim (1968), have defined career stage by the number of years in an occupation or by the length of tenure in an organization: establishment (less than 2 years tenure), advancement (2 to 10 years tenure), and maintenance (more than 10 years tenure). This definition of career stage is less likely to classify inaccurately people with irregular career patterns (e.g., women returning to the paid labour force after taking time out for family responsibilities), and was therefore the career stage model that was utilized in the present study. During the establishment phase, the individual is a newcomer to his or her occupation or organization. The major concern during this phase is to become established within the organization (Hall, 1976). Establishment requires the development of competence and gaining the acceptance of peers (Feldman, 1976; Schein, 1978). In the advancement stage, the key concerns are with increasing responsibility, upward mobility, and promotions (Hall & Nougaim, 1968). During the maintenance stage, the individual focuses on supporting the development of other organizational members and the organization as a whole (Hall & Nougaim, 1968). The individual's needs for mentoring relationships will be greatest, then, during the establishment and advancement stages.  F. RESEARCH QUESTIONS AND HYPOTHESES The hypotheses which emerged from the preceding review of the literature are presented in this section. Each of the five research questions that were originally stated in Chapter I (Introduction) are restated prior to the listing of the hypotheses which correspond to that particular question.  78 Research Question 1: What is the relationship between an individual's homosocial attitudes and the size of his or her same- and cross-gender networks?  The three homosocial attitudes common to women and men (Cross-Gender Preference, Same-Gender Preference, and Valuing) and the two women-only homosocial attitudes (Identification and Jealousy) are referred to in the two hypotheses which correspond to Research Question 1.  Hypotheses Hypothesis 1. The same-gender networks of respondents who report low levels of Jealousy (women only) and Cross-Gender Preference and high levels of Identification (women only), Same-Gender Preference, and Valuing will be significantly larger than the same-gender networks of respondents who report high levels of Jealousy (women only) and Cross-Gender Preference and low levels of Identification (women only), Same-Gender Preference, and Valuing.  Hypothesis 2. The cross-gender networks of respondents who report high levels of Jealousy (women only) and Cross-Gender Preference and low levels of Identification (women only), Same-Gender Preference, and Valuing will be significantly larger than the cross-gender networks of respondents who report low levels of Jealousy (women only) and Cross-Gender Preference and high levels of Identification (women only), Same-Gender Preference, and Valuing.  These hypotheses arose from the theories and research related to homosocial attitudes which were discussed in the Homosocial Attitudes section of this chapter. Examples of how homosocial attitudes may have affected the results of other studies discussed in this chapter were interspersed throughout the Homosociality and Organizational Networks and Mentoring sections as well. The reader may refer to these sections for further explanation of the rationale for these hypotheses.  79 The basic idea behind Hypothesis 1 is that if people have favourable attitudes toward samegender persons, they will seek relationships with them and, hence, list more same-gender persons in their networks. Hypothesis 2 predicts that people who have unfavourable attitudes toward their own gender, perhaps preferring cross-gender relationships instead, will seek more relationships with cross-gender persons and, hence, will list more cross-gender persons in their networks.  Research Question 2: What is the relationship between an individual's homosocial attitudes and the levels of supportive (mentoring) behaviours and relationship provisions he or she receives from same- and cross-gender networks?  The same homosocial attitudes as those referred to in the hypotheses corresponding to Research Question 1 are referred to in the hypotheses corresponding to Research Question 2. The overall level of supportive (mentoring) behaviours and the levels of six different relationship provisions (Intimacy, Similarity, Defiance of Convention, Emotional Support, Respect for Differences, and Lifetime Attachment) received from same- and cross-gender networks are addressed by the two hypotheses which correspond to Research Question 2. For the sake of simplicity, the relationship provisions are not listed separately in the hypotheses.  Hypotheses Hypothesis 3. Respondents who report low levels of Jealousy (women only) and CrossGender Preference and high levels of Identification (women only), Same-Gender Preference, and Valuing will receive significantly higher levels of supportive (mentoring) behaviours and relationship provisions from same-gender networks than will respondents who report high levels of Jealousy (women only) and Cross-Gender Preference and low levels of Identification (women only), SameGender Preference, and Valuing.  80 Hypothesis 4. Respondents who report high levels of Jealousy (women only) and CrossGender Preference and low levels of Identification (women only), Same-Gender Preference, and Valuing will receive significantly higher levels of supportive (mentoring) behaviours and relationship provisions from cross-gender networks than will respondents who report low levels of Jealousy (women only) and Cross-Gender Preference and high levels of Identification (women only), SameGender Preference, and Valuing.  The rationale for these hypotheses was similar to that expressed for Hypotheses 1 and 2. However, Hypotheses 3 and 4 differ from the first two: Hypotheses 1 and 2 examined the relationship between homosocial attitudes and network size, while Hypotheses 3 and 4 addressed the relationship between homosocial attitudes and network activities (mentoring behaviours and relationship provisions). The gist of Hypothesis 3 is that persons with favourable attitudes toward their own gender are more likely to receive higher levels of supportive (mentoring) behaviours and relationship provisions from same-gender persons than are persons with unfavourable attitudes toward their own gender. Hypothesis 4 predicts that persons with unfavourable attitudes toward their own gender, perhaps preferring cross-gender relationships instead, are more likely to receive higher levels of mentoring behaviours and relationship provisions from cross-gender persons than from same-gender persons. Persons with the pattern of homosocial attitudes described in Hypothesis 4 might be expected to reject any offers by same-gender persons to provide mentoring behaviours and relationship provisions.  Research Question 3: Are there gender differences in the levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks?  The same supportive (mentoring) behaviours and relationship provisions as those referred to in the hypotheses listed under Research Question 2 are referred to in the hypotheses which correspond to Research Question 3.  81  Hypotheses Regarding Supportive (Mentoring) Behaviours Hypothesis 5. Same-gender networks will provide significantly higher levels of supportive (mentoring) behaviours than will cross-gender networks.  Larwood and Blackmore's (1978) findings that both men and women were more likely to seek same-gender than cross-gender acquaintances for leadership tasks suggested that Hypothesis 5 would be supported. The findings of Henderson (1985), that "women executives were three times more likely than a man to have a woman as a mentor" (p. 860) and of Reich (1986), that 9 0 % of the women in his study who had become mentors were mentors to other women, while fewer than 5 % of the men studied had women proteges, also supported this hypothesis. The findings of Brass (1985), that gender-segregated informal networks appeared to exist in the organization he studied, also suggested that same-gender persons were more likely to provide higher levels of mentoring behaviours than were cross-gender persons. Hypothesis 6 (which follows) was formulated in partial opposition to Hypothesis 5. The rationale for the formulation of an oppositional hypothesis is as follows. In a male-dominated organization, women who desire career advancement may need to seek mentoring behaviours from men (who occupy positions of greater power within the hierarchy) rather than from other women (who may be present in small numbers in higher level positions) (Hennig & Jardim, 1977; Noe, 1988). While respondents for the proposed study were drawn from work groups with as close to a balanced gender ratio (Kanter, 1977) as possible, the upper echelons of the majority of organizations are still occupied mainly by men. Women seeking significant career advancement still may choose primarily cross-gender relationships to assist in this pursuit, since other women in the work setting may not be in positions which would enable them to supply high levels of mentoring behaviours. Women seeking the acquisition of power in a male-dominated system may do this by forming relationships with (powerful) men; other women, because of their relative lack of power in  82 this environment, may be perceived as unhelpful in gaining increasing control over resources (Lipman-Blumen, 1976). Hence, Hypothesis 6:  Hypothesis 6. In partial opposition to Hypothesis 5, it is predicted that women's crossgender networks will provide significantly higher levels of supportive (mentoring) behaviours than will women's same-gender networks.  Finally, the theories of Lipman-Blumen (1976) and Tiger (1969) suggested Hypothesis 7:  Hypothesis 7. Men's same-gender networks will provide significantly higher levels of supportive (mentoring) behaviours than will women's same-gender networks.  Both Lipman-Blumen and Tiger state that the basis of male bonding is an attraction to power and status, that men bond around the acquisition of resources and exclude women from these bonds. Women traditionally have not had access to the same level of power and resources as have men. Hence, unlike men, women have not been as able to offer each other significant assistance in acquiring resources (remember that in a mentoring relationship, the protege learns how to navigate the organization and to gain control over an increasing number of resources). Hypothesis 7 was also supported by the findings of Lincoln and Miller (1979), that women and nonwhites were excluded from organizational networks that were a medium for the attainment of organizational goals. Brass' (1985) findings also lent support to Hypothesis 7. In his study, promotions and influence were significantly related to centrality in the men's informal network and to the informal network of the dominant coalition of men. The women in Brass' (1985) study were not as central to these networks as were the men.  83 Hypotheses Regarding Relationship Provisions Hypothesis 8. The same-gender networks of women and the cross-gender networks of men will provide significantly higher levels of Intimacy and Emotional Support than will the samegender networks of men and the cross-gender networks of women.  Hypothesis 8 was suggested by the theories of Pleck (1975) and Safilios-Rothschild (1981), and by the research findings of Arnold and Chartier (1986), Toews (1973), and Wetstein Kroft (1987/1988), all of which indicate that both women and men tend to have needs for intimacy and emotional closeness met primarily in relationships with women. The findings of Burke (1984a), Collins (1983), and Reich (1986) - that mentoring relationships involving women as mentor or protege are richer in the psychosocial dimension than those of men - also supported this hypothesis.  Hypothesis 9. Men's same-gender networks will provide significantly higher levels of Respect for Differences than will women's same-gender networks.  Hypothesis 9 was suggested by the work of Eichenbaum and Orbach (1983a, 1983b, 1988) who contend that central to women's psychology is a difficulty establishing boundaries between the self and others. They state that women's tendency toward merged attachments can be traced to the dynamics of the mother-daughter relationship in which generational boundary crossing and its concomitant betrayal of the child's trust are common. Women learn to be the emotional caregivers in this key relationship with their mothers and, failing to establish a clear distinction between the mother and the self (as do male children), thereafter tend to merge with others in relationships. Put in very simple terms, "differentiation defies the very essence of feminine psychology" (Eichenbaum & Orbach, 1988, p. 50).  84 Hypothesis 10. The same-gender networks of both women and men will provide significantly higher levels of Similarity than will the cross-gender networks of women and men.  Hypothesis 10 emerged from the work of Arnold and Chartier (1986), Larwood and Blackmore (1978), Lipman-Blumen (1976), and Wetstein Kroft (1987/1988). As "ingroup" members, same-gender persons are more likely to be perceived as being similar to the respondent in terms of values, interests, and activity preferences than are "outgroup" members (cross-gender persons).  Hypothesis 11. Men's same-gender networks will provide significantly higher levels of Lifetime Attachment than will women's same-gender networks.  Hypothesis 11 was based on Tiger's (1969) theory that men's same-gender relationships are stronger and more enduring than are women's same-gender relationships. While Tiger's theory is an older one and has been contradicted by research findings on relationships in non-work settings (Booth, 1972; Smith-Rosenberg, 1975; Woolsey & McBain, 1987b), there are no known empirical investigations of his theory based on work relationships. Given the nature of his theory (i.e., that males form lasting bonds with each other in hierarchical, power-based structures), it seemed important to test this theory in an organizational setting. The contradictory findings of Booth (1972), Smith-Rosenberg (1975), and Woolsey and McBain (1987b) suggested the formulation of an oppositional hypothesis to Hypothesis 11. Booth (1972) found that women were as likely as men to show stable patterns of participation in samegender voluntary organizations as were men. Smith-Rosenberg (1975) documented emotionally rich and long-lasting bonds between women in the nineteenth century. The Woolsey and McBain (1987b) interviews with women and men about their same-gender relationships indicated that both women and men formed long-lasting same-gender social bonds. Hence, Hypothesis 12 was formulated as an oppositional hypothesis to Hypothesis 11:  85 Hypothesis 12. In opposition to Hypothesis 11, it is predicted that there will be no significant difference in the levels of Lifetime Attachment provided by women's and men's samegender networks.  Some respondents in Woolsey's (1984) interview study stated that same-gender relationships were longer-lasting than were cross-gender relationships because same-gender relationships were not complicated by the issue of sexuality. Other writers and researchers (e.g., Clawson & Kram, 1984; Kram, 1985; Reich, 1986), in discussing the difficulties of cross-gender mentoring, have noted that issues such as sexual attraction and unwanted sexual advances have caused relationship difficulties, some of which ended in a termination of the relationship. It seemed, then, that because same-gender relationships were less likely to be complicated with the issue of sexuality, that same-gender relationships in the workplace would have higher levels of Lifetime Attachment than would cross-gender relationships. Hence, Hypothesis 13 reads:  Hypothesis 13. The same-gender networks of women and men will provide significantly higher levels of Lifetime Attachment than will the cross-gender networks of women and men.  Research Question 4: How do homosocial attitudes in interaction with levels of supportive (mentoring) behaviours received from same- and cross-gender networks relate to careerand job-related outcomes?  An explanation of how the data were handled for this question is necessary prior to stating the hypotheses. A number of data reduction techniques were employed to streamline the analyses utilized in testing the two hypotheses formulated for this research question while allowing maximum use of the available data. These data reduction analyses are described in more detail in Appendix C (Preliminary Analyses). For now, a brief explanation is provided in order to clarify the meaning of Hypotheses 14 and 15.  86 First, in order to derive composite measures of homosocial attitudes, factor analyses were performed on the correlation matrices for the three (for men) and five (for women) homosocial attitudes scales. These analyses produced two composite factors which were conceptually comparable for women and men. Factor scores on these two composite factors -- Same-Gender Preference Composite and Cross-Gender Preference Composite - are the scores referred to in Hypotheses 14 and 15. Second, the scoring for the measure of supportive (mentoring) behaviours utilized in this study (Schroeder, 1988/1989) was modified somewhat for this research question. Respondents who listed no same-gender providers of supportive (mentoring) behaviours were given a score of zero for level of mentoring received from same-gender networks; respondents who listed no crossgender providers of supportive behaviours received a score of zero for level of mentoring received from cross-gender networks. Scores were computed in the usual manner for respondents who had listed one or more same-gender mentors and had completed the mentoring measure with respect to the level of mentoring behaviours provided by these same-gender persons (i.e., the score for the level of same-gender mentoring received was the actual score received on the measure and was indicative of the overall level of mentoring received from same-gender persons). Scores also were computed in the usual manner for persons who had completed the mentoring measure with respect to one or more cross-gender providers of supportive (mentoring) behaviours. The career- and job-related outcomes referred to in Hypotheses 14 and 15 were introduced in the Mentoring section of this chapter. Scores on 10 separate measures, broadly defined as career- and job-related outcomes, were factor analyzed and a three-factor structure conceptually comparable to that found by Schroeder (1988/1989) was obtained. Factor scores on each of the three career- and job-related factors - named Affective, Advancement, and Skill Development Outcomes in the present study -- are the dependent variables referred to in Hypotheses 14 and 15.  87 Hypotheses Hypothesis 14. Respondents with high scores on the Same-Gender Preference Composite who receive high levels of mentoring from same-gender networks will have significantly higher levels of Affective, Advancement, and Skill Development Outcomes than respondents with high scores on the Same-Gender Preference Composite who receive low levels of mentoring from same-gender networks.  Hypothesis 15. Respondents with high scores on the Cross-Gender Preference Composite who receive high levels of mentoring from cross-gender networks will have significantly higher levels of Affective, Advancement, and Skill Development Outcomes than respondents with high scores on the Cross-Gender Preference Composite who receive low levels of mentoring from cross-gender networks.  Question 4 and Hypotheses 14 and 15 address the issue of availability of appropriate sameor cross-gender persons in the respondents work setting. An individual who indicates a strong preference for associating with same-gender persons (i.e., receives a high score on the SameGender Preference Composite) but finds that like-minded same-gender persons are not available in his or her environment, may not fare as well as someone who does not particularly care to associate with same-gender persons. Take, for example, the case of a woman who receives a high score on the Same-Gender Preference Composite and who finds herself in a work setting with other women who score low on the Same-Gender Preference Composite. The other women in her environment who "invidiously disesteem ... the qualities associated with women" (Laws, 1975, p. 54) may cultivate relationships with men in the work setting and avoid relationships with women. The lone woman in this setting who seeks and prefers relationships with other women might experience a relational deficit (i.e, she might receive a low level of same-gender mentoring). This deficit may manifest itself as low levels of career- and job-related outcomes. The same principle might hold for  88 someone who strongly prefers cross-gender relationships (high score on the Cross-Gender Preference Composite) and who receives a low level of cross-gender mentoring.  Research Question 5: How do the levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks relate to careerand job-related outcomes?  The usual method of scoring the supportive (mentoring) behaviours measure (Schroeder, 1988/1989) was employed in the analyses for Research Question 5. Six relationship provisions were measured by the Close Social Relationships Scale (Woolsey et al., 1988b): Intimacy, Similarity, Defiance of Convention, Emotional Support, Respect for Differences, and Lifetime Attachment. The career- and job-related outcomes were the same as those described in the preceding section (Research Question 4).  Hypotheses Hypothesis 16. For both women and men, there will be significant positive relationships between the levels of supportive (mentoring) behaviours and relationship provisions received from any source (same- or cross-gender network) and Affective, Advancement, and Skill Development Outcomes.  An increasing number of writers and researchers attest to the link between having a mentor or receiving mentoring behaviours on the one hand and enhanced career and personal development on the other (Collins & Scott, 1978; Noe, 1988; Reich, 1985,1986; Schroeder, 1988/1989). Given that Schroeder (1988/1989) found significant positive relationships between the level of supportive (mentoring) behaviours received and each of three composite career- and job-related outcomes (promotional, skill development, and job-related), the current study was likely to find the same. While the individual outcome measures and the method of completing the supportive (mentoring)  89 behaviours measure in the present study varied somewhat from the measures and methods employed by Schroeder (1988/1989), the same general pattern of results was expected to hold. Hence, the formulation of Hypothesis 16. Finally, because relationship provisions, as measured by the Close Social Relationships Scale (Woolsey et al., 1988b), were expected to tap more deeply into the character of the emotional bond that can be present in supportive work relationships, it was expected that these provisions would be more strongly related to respondents' affective responses than to their advancement and skill development. Hence, Hypothesis 17 reads:  Hypothesis 17. For both women and men, the level of relationship provisions received from any source (same- or cross-gender network) will be significantly more strongly related to Affective Outcomes than to Advancement and Skill Development Outcomes.  This hypothesis is an important one for women in particular. Kanter (1977) has suggested that persons who perceive few opportunities for career advancement will direct their energy into developing satisfying personal relationships at work rather than focusing on career advancement. Research has shown that women's relationships are richer in the psychosocial domain than are men's (Caldwell & Peplau, 1982; Reich, 1986; Reis, Senchak, & Solomon, 1985). A woman could be receiving high levels of relationship provisions.and, hence (if the predicted relationship holds true), score high on the Affective Outcomes factor. This same woman could conceivably be receiving low levels of mentoring behaviours and score low on the Advancement and Skill Development Outcomes factors.  90 CHAPTER III. METHOD  The method of data collection selected for this study was a questionnaire that was completed by individual staff members in seven different organizations. The research procedures -recruitment of participating organizations, generation of the sample, and questionnaire distribution and return - are presented in the first section of this chapter. Following that, the measuring instruments, design, and statistical analysis are discussed.  A. RESEARCH PROCEDURES  1. Recruitment of Participating Organizations Senior members (e.g., senior personnel department representatives) of 32 organizations in Western and Central Canada were contacted by telephone or by letter and a follow-up telephone call to invite the participation of each organization in the study. Those contact people indicating an interest in the study were provided a written summary of the nature of. the study, a list of the selection criteria for potential respondents, and sample questionnaires. Each representative was informed that written feedback summarizing the results of the study would be provided to the organization and to participating staff members who requested it. The investigator discussed the selection criteria and methods (outlined in the next subsection) to be used in generating the sample of potential respondents with a senior member of each interested organization (usually a senior personnel department representative) to be sure that an appropriate sample could be generated from the organization. A total of seven organizations, all located in large Western Canadian centres, participated in the study: two were from the public sector (both post-secondary education), four were from the private sector (computer systems design and consultation, equipment sales and service, professional accounting, and telecommunications broadcasting), and one was a Crown corporation (utilities).  91  2. Generation of the Sample Four criteria were applied in selecting potential respondents in each organization. First, respondents were to be employed in managerial, supervisory, professional, or technical positions, as this was the population of interest for this study. Second, because this was a study of a variable (homosociality) hypothesized to be related to organizational socialization and career advancement, it was important to sample from a pool of employees who were in relatively early stages of career development or in the early years of organizational tenure. Therefore, respondents were limited to those with 15 years or less of organizational tenure. Third, because "physical proximity acts as a necessary precondition for the formation of relationships, that is, people have to have contact with each other to form relationships" (Woolsey & McBain, 1987b, p. 62), respondents had to be employed in gender-integrated work groups (approximately 50:50 to 75:25 gender ratio), thereby ensuring that they had the opportunity to form relationships with people of both genders. Finally, because relatively equal and large numbers of questionnaires completed by women and men (approximately 250 each) were required for the proposed data analyses, and because the response rate for men in relationship research tends to be lower than that for women (Woolsey et al., 1988a, 1988b), the number of male potential respondents was to be equal to or somewhat greater than, the number of female potential respondents. In one post-secondary education institution, the pool from which the sample was drawn consisted of (a) all full-time managerial, supervisory, and professional staff members classified within one large employment group, and (b) all full-time technical staff members above a certain job level. The sample consisted of all 1,112 people (545 women, 567 men) in this pool who met the tenure and gender-integrated work group criteria. In the other post-secondary education institution, the sample was drawn from a pool of managerial, supervisory, and professional staff members classified within one large employment group. A list of 232 people (122 women, 110 men) who met the tenure and gender-integrated work group requirements was generated; all 232 people were included in the sample.  92  The sample from the computer systems design and consultation firm was obtained by selecting professional and technical staff members who met the tenure and gender-integration requirements. This group included 8 women and 9 men, 17 people in total. All managerial, supervisory, and professional staff members who met the tenure and genderintegrated work group requirements were included in the sample from the equipment sales and service company. The 10 people in this sample included 6 women and 4 men. The sample from the professional accounting firm consisted of 20 managers (10 women, 10 men) who met the tenure and gender-integrated work group requirements. The sample from the organization in the telecommunications broadcasting industry was derived by listing all supervisors and managers who met the tenure and gender-integrated work group criteria. The list of 38 people contained the names of 21 women and 17 men. The sample from the corporation in the utilities industry was generated by selecting all managers, supervisors, and professional staff who met the tenure and gender-integration requirements. The names of 27 people (13 women, 14 men) were provided to the investigator. In summary, the total sample consisted of 1,456 staff members (725 women, 731 men) employed in one of seven organizations located in large Western Canadian centres.  3. Questionnaire Distribution and Return The questionnaire was divided into two parts, each with a separate administration, in order to reduce each respondent's time commitment for each sitting to approximately 30 minutes. A small pilot study was undertaken prior to the main study, primarily to test clarity of instructions and length of time required to complete each part of the questionnaire. Each of 12 respondents (5 women, 7 men) employed in administrative, professional, or technical positions in different organizations in the Lower Mainland of British Columbia completed one version of progressively different formats of the two-part questionnaire. Verbal feedback received from these pilot respondents and information gained by visual inspection of their completed questionnaires were used to revise the questionnaire prior to distributing it in the first organization.  93 Each of the 1,456 staff members in the sample was mailed an envelope containing a cover letter from a senior personnel representative, a cover letter from the author, the first part of the questionnaire, and a reply envelope in which to return the completed questionnaire directly to the author, and, in the case of six organizations in which data were collected, her thesis advisor. In the case of one organization in which data were collected (post-secondary education), reply cards for requesting feedback also were included. The cover letter from each personnel representative indicated to the staff members that the organization's management endorsed the study. The cover letter from the author summarized the study, assured anonymity and confidentiality, and indicated that feedback about the results of the study would be provided to interested participants upon request. Within one week of distributing Part One of the questionnaire, an envelope containing Part Two and a return envelope was mailed by the investigator to each of the potential respondents to whom Part One had been mailed. Considerable difficulty was experienced at this stage in one organization in which data were collected (post-secondary education). Within days of the questionnaires being distributed, the author was contacted by staff members who wished to offer feedback on behalf of individuals or groups of concerned colleagues who had received questionnaires. Those persons giving the feedback described a generalized suspicion of the larger management structure in their organization; participation in the current study was perceived by some people as a threat to an already shaky sense of job security. The feedback offered by concerned staff members indicated that three aspects of the questionnaire package had aggravated this suspicious response. First, because the questionnaire had been accompanied by a cover letter from a senior member of the organization, the perception of some people was that the author had been hired by the organization's management to conduct the study (rather than the survey being undertaken independently for her doctoral dissertation research). Second, in two sections of the questionnaire, respondents were requested to list the initials, gender, age, and relationship (to the respondent) of persons in their organization who  94 provided them with certain mentoring behaviours or relationship provisions. Some people were concerned that this was too much identifying information to provide; they did not like the feeling that they were naming colleagues and reporting on them. Third, for questionnaire matching purposes, each set of questionnaires (Part One and Part Two) had been stamped with a code number just prior to inserting the parts into separate envelopes, both of which had been pre-labelled with one potential respondent's name and address. As Parts One and Two of the questionnaire were to be completed and returned separately, these code numbers were to be used to match the two parts of the questionnaire upon their return to the author. While no records had been kept to link particular code numbers to individuals, some people indicated a fear that their responses would not remain anonymous. The feedback noted above and a much lower than expected response rate prompted the early distribution of a follow-up letter from the author to all potential respondents only two days after mailing Part Two of the questionnaire; a second follow-up letter from the investigator's Supervisory Committee was mailed 12 days later. Both letters thanked staff members for their participation and reiterated (a) that the study was being undertaken as doctoral dissertation research, and (b) that all responses were anonymous and confidential. For the benefit of those people who were particularly concerned about the code numbers stamped on the questionnaires, instructions were given for returning unnumbered questionnaires. In addition, the second follow-up letter provided instructions for using code letters rather than initials to identify co-workers who were providing the behaviours of interest. Little improvement in response rate was noted after the follow-up letters. While it was believed that the low response rate in this organization was due in large part to the suspicious organizational climate, some modifications aimed at reducing concerns about anonymity were made to the questionnaire before distributing it in the other six organizations. First, potential respondents were requested to provide their own questionnaire matching codes. Each respondent was asked to make up and keep a record of a six-character code, to print the code on Part One of the questionnaire prior to returning it, and to place the same code on Part Two. for the benefit of people who preferred not to identify co-workers by their initials, optional  Second,  95 instructions for using code letters were provided. Other minor changes made to the questionnaire prior to distributing it in the final six organizations included highlighting important instructions that were missed by some people in the first organization and deleting one demographic item regarding current interest in developing a new romantic relationship that may have contributed to a lowering of response rate. This item was of minor importance in the proposed analyses and so it was deleted. Data collection in the final six organizations proceeded uneventfully. Follow-up letters were distributed to all potential respondents two weeks after Part Two of the questionnaire was mailed. These letters thanked participants who had already returned both parts of the questionnaire and requested those who had not to do so soon. Questionnaires were distributed in the seven organizations over a seven-month period and returned over an eight-and-one-half-month period. The return rate is presented in Chapter IV (Results).  B. MEASURES The measuring instruments used in this study are described in this section. The instructions and items for each of the measures are located in Appendix A. The measures appear in Appendix A in the order in which they appeared in the actual questionnaire. For ease of administration, the questionnaires were colour-coded according to the gender of the potential respondent (blue for women, grey for men). The only difference between the women's and the men's questionnaires was that the women's form contained the Bonds Between Women Scale in Part Two and the men's form contained the Bonds Between Men Scale. To guard against response set bias, the response format was counterbalanced for those items requiring separate responses for same- and cross-gender networks. For example, in half of the questionnaires distributed to women, respondents were requested to answer the supportive behaviour items and the Close Social Relationships Scale items first while considering their relationships with women in the organization, and second while considering their relationships with men in the organization. In the other half of questionnaires distributed to women, respondents were  96 requested to answer first for relationships with men and second for relationships with women. The same procedure was employed with questionnaires distributed to men. A description of each of the measures contained in the questionnaire follows.  1. The Close Social Relationships Scale (CSRS) Development of the Scale. The Close Social Relationships Scale (Woolsey et al., 1988b) was designed to measure homosociality at the level of actual relationships. That is, scale items were designed to measure respondents' thoughts, feelings, and behaviours in regard to specific, lived relationships. Unless noted otherwise, the descriptions of CSRS scale development summarized in this section have been extracted from the, as yet, unpublished data analyses of Woolsey et al. (1988b). Scale construction began with interviews of 35 women and 35 men on the topic of their same-gender social relationships with friends, relatives, co-workers, and neighbours. The sample for this interview study was representative of a university-educated population of adult women and men: 6 9 % of the women and 80% of the men were university students; all but one were between the ages of 18 and 35. The sample was varied in terms of marital status and the presence of children (parental status); as well, over 28% of the sample was of a self-designated ethnic origin other than Canadian (Woolsey & McBain, 1987b). Items for the original version of the CSRS were constructed primarily on the basis of the friendship data. Critical incidents reported to have strengthened same-gender friendship bonds were organized into themes or categories; the model of friendship bonds that emerged from this categorization showed "a progression in friendship from bonds of a casual nature to bonds of a deeper, more enduring nature" (Woolsey & McBain, 1987b, p. 63). Scale items were written to reflect both the prototypical incidents and the general themes described in this model. The original 187-item version of the CSRS was completed by 1,226 respondents. Of these, 961 (645 women, 316 men) were university or community college students; the remainder (163 women, 102 men) were recruited from non-credit community adult education classes. The student  97 and community samples differed in terms of age and marital status, with the student samples tending to be younger (the mean ages were 27.40 years for women and 25.20 years for men) and predominantly single (55.3% of the women and 64.2% of the men were single) and the community samples tending to be older (the mean ages were 37.57 years for women and 40.41 years for men) and predominantly married (63.8% of the women and 74.5% of the men were married). In preparation for completing the scale, respondents were requested to name up to 10 same-gender persons with whom they had their closest social relationships. They were then asked to indicate how strongly they believed each item to be true or not true of the relationships they had listed. The 6-point scale for responses ranged from 1 (definitely not true) to 6 (definitely true). Principal component analyses were performed separately on the data gathered from each of the four different groups: student women (n = 645), student men (n = 316), community women (n = 163), and community men (n = 102). The data from both community samples were eventually discarded because of small sample sizes and poor matching of factor structures and item factor loadings with the stronger results for the two student samples. The unclear results for the community samples and the subsequent discarding of these data were unfortunate because these community samples were, in some respects (such as age, marital status, and occupational groupings), more similar to the target group for the current study than were the student samples. It would have been better for the purpose of the current study, to have used a measure of relationships developed from data gathered from respondents at a more similar point in the lifecycle to the potential respondents in the current study. However, such a measure was not available. The results of an oblique 10-factor solution (using principal component analysis with the student women and student men data) provided the most conceptually meaningful solution with the greatest number of item matches across the student women and student men samples. The item content of 8 of the 10 factors reflected the majority of the general themes or categories outlined in the model of friendship bonds developed from the critical incident study described earlier (Woolsey & McBain, 1987b). The 8 interpretable factors were named: Casual Socializing, Similarity, Defiance  98 of Convention, Emotional Support, Respect for Differences, Practical Support, Intimacy, and Lifetime Attachment. The 2 remaining factors were not interpretable. Eighty-six items were retained for further analyses, based on factorial simplicity of matched items and conceptual meaning of the factors. In most cases, an item was retained only if: (a) it had a factor loading of .30 or more for both the student women and student men samples, (b) it was factorially simple (i.e., did not have a factor loading of .30 or greater on more than one factor), and (c) it had a factor loading of .30 or more on one of the eight conceptually meaningful factors of the 10-factor solution. The retained items were re-factored twice, using the original data from the student samples. Eight-factor unweighted least squares factor analyses, employing Harris-Kaiser oblique transformations, were performed on the correlation matrices for 86, then 59, retained items. Once again, items were retained or discarded according to the decision rules outlined above. Fifty-four items representing eight factors were retained after the final round of factor analyses with 59 items. All of the retained 54 items were factorially simple, remained on their original factors, and, with only two exceptions with factor loadings of .29, had pattern coefficients greater than or equal to .30 for both the student men and student women samples. Only six of the eight factors were used in the current study. Practical Support and Casual Socializing were excluded because they were less important (conceptually) to the study. Deletion of the items on these subscales brought the number of items to be used in the current study down to 47. Interscale correlations for the 47-item version of the CSRS used in the current study were acceptably low, ranging from .10 to .60 for the student women and student men samples. The addition of three new positively worded items to the Intimacy subscale brought the total number of CSRS items used in the current study up to 50. These three new positively scored items were added because all of the other Intimacy items were reverse scored. The 50 CSRS items utilized in the current study are listed in Part Two of the questionnaire under the Close Social Relationships Scale (see Appendix A). The numbers of the items corresponding to each of the six subscales are (new items in bold print and reverse-scored items underlined): Similarity (1, 7,13, 19, 25, 31, 36, 40,  99 45, 48); Defiance of Convention (2, 8, 14, 20, 26, 32, 37, 42, 46, 49); Intimacy (3, 9,15, 21, 27,30, 33, 38,41, 43, 47,50); Emotional Support (4,10,16, 22, 28, 34, 39, 44); Respect for Differences (5,11, 17, 23, 35); and Lifetime Attachment (6, 12, 18, 24, 29). Other changes made to the CSRS to adapt it for use in the current study include the following. First, the method of generating the initials of network members in preparation for completing the scale was altered to include elements of procedures used for similar purposes by Fischer (1982) and Schroeder (1988/1989). Respondents in the current study were requested to list the initials of up to five persons (gender not specified) who provided each of the following six relational provisions, listed respectively: Similarity, Emotional Support, Respect for Differences, Lifetime Attachment, Intimacy, and Defiance of Convention. The six questions (items 1-6) used to generate these initials were placed in the introductory section of the Close Social Relationships Scale in Part Two of the questionnaire (see Appendix A). Second, following this listing of initials, respondents were asked to transfer into a Network Table the initials of up to 10 people listed most frequently in response to the six relationship provisions questions. The Network Table included several questions designed to elicit descriptive information about each of the relationships listed in the table. Finally, respondents were instructed to respond to each of the 50 CSRS items while considering (a) the same-gender people listed in the CSRS Network Table who also had been identified as providers of supportive (mentoring) functions in the first part of the questionnaire, and (b) the cross-gender people identified in the CSRS Network Table who also had been cited as providers of supportive (mentoring) functions in Part One. The CSRS Network Table and instructions for completing the 50 CSRS items while considering each of the networks are located under the title Close Social Relationships Scale in Part Two of the questionnaire (see Appendix A).  Scoring. Each item on the CSRS receives a score from 1 (definitely not true) to 6 (definitely true). A separate score is derived for each of the six subscales by reversing the value of negatively worded items and then summing the responses to each item contained in the subscale.  100 Reliability. Internal consistency coefficients (alpha) for each of the six subscales in the 47item CSRS used in the current study were computed from the student women and men data. Most of the coefficients were above .80. The lowest were .58 (for student men) and .64 (for student women) on Respect for Differences. Given that this was only a 5-item subscale, this level of internal consistency was judged to be adequate. Test-retest reliability has not been investigated.  Validity. Convergent and discriminant validity of the CSRS had not been assessed prior to the time of data collection for the present study. Construct validity was assessed, in part, by the repeated factor analyses (Woolsey et al., 1988b) described above, in which the content and meaning of the eight factors remained consistent. Factor analyses performed on the new data collected for the current study continued this process (these results are presented in Appendix C [Preliminary Analyses]). The content of the eight factors found by Woolsey et al. (1988b) also was compared with the content of other measures of friendship and intimacy available prior to the time of data collection for the current study. The eight CSRS factors were found to contain almost all of the content of the following nine measures taken in combination: Relationship Rating Form (Davis & Todd, 1985), Acquaintance Description Form (Wright, 1985), Friendship Expectancy Inventory (La, Gaipa, 1977), Intimacy Categories Scale (Arnold & Chartier, 1986), Solidarity Scale (Arnold & Chartier, 1986), Emotional Intimacy Scale (Williams, 1985), Miller Social Intimacy Scale (Miller & Lefcourt, 1982), Love-Scale and Liking-Scale (Rubin, 1970), and the Social Provisions Scale (Cutrona & Russell, 1987; Russell & Cutrona, 1984). In addition, the CSRS had a factor (Defiance of Convention) that was not found in any other measure. Among the most comprehensive of the other measures - the Relationship Rating Form (Davis & Todd, 1985), the Acquaintance Description Form (Wright, 1985), the Friendship Expectancy Inventory (FEI) (La Gaipa, 1977), the Intimacy Categories and Solidarity Scales (Arnold & Chartier, 1986), and the Social Provisions Scale (Cutrona & Russell, 1987; Russell & Cutrona, 1984) - scale development had not proceeded to the point of establishing the same simple factor structure for both genders and for both same- and cross-gender relationships. As a  101 major purpose of the current study was to compare women's and men's same- and cross-gender relationships, a comprehensive measure with an established simple factor structure valid for use with all of these relationships was sought. The CSRS had the advantage of the repeated factor analyses described earlier in which the same simple factor structure remained consistent for both student women and student men. One drawback of using the CSRS to study cross-gender relationships was that the CSRS was developed from interview data about same-gender relationships; in addition, respondents to the original 187-item version of the CSRS were requested to complete the scale with their closest samegender relationships in mind. However, it should be noted that La Gaipa (1977) wrote the items for the FEI from interview data about same-gender relationships too (and the FEI subscales match the CSRS factors very closely). The FEI subsequently has been used to investigate both same- and cross-gender relationships. In addition, the majority of the other nine scales reviewed have been used to investigate both same- and cross-gender relationships. For example, the Intimacy Categories Scale and Solidarity Scale (Arnold & Chattier, 1986) were completed twice by 143 single undergraduate men: once for their closest same-gender relationship and once for their closest cross-gender relationship. Common factor analysis with varimax rotation resulted in similar factor structures for both same- and cross-gender relationships. The data analysis for the current study included performing separate factor analyses on the CSRS data gathered from women and men in regard to their same- and cross-gender work relationships. The results of these analyses, reported in Appendix C (Preliminary Analyses), supported the validity of using the CSRS to measure aspects of women's and men's same- and cross-gender relationships.  2. The Bonds Between Women/Bonds Between Men Scale (Bonds) Development of the Scale. The Bonds Between Women/Bonds Between Men Scale (Woolsey et al., 1988a) was designed to measure homosocial attitudes, that is, attitudes toward same-gender persons in general (i.e., toward one's own gender as a group). The scale has two  102 forms. The Bonds Between Women Scale was designed to be completed by women, and the Bonds Between Men Scale was designed to be completed by men. The two forms are different in two respects. First, the men's form has three factors or subscales while the women's form has five. Three factors -- Cross-Gender Preference, SameGender Preference, and Valuing ~ are common to women and men. Two factors ~ Identification and Jealousy - are women-only factors. The separate factors for women were the result of repeated conceptual and factor analyses, which are described later in this subsection. The second difference between the women's and men's forms is that the words "women" and "men" have been interchanged as appropriate in all of the items common to women and men. For example, in the Bonds Between Women Scale, an item reads, "My relationships with women are very important to me," while in the Bonds Between Men Scale, the same item is worded, "My relationships with men are very important to me." The 192 items on the original questionnaire were drawn from several sources. First, items were written on the basis of interview data obtained from 35 men and 35 women on the topic of their same-gender social relationships (Woolsey & McBain, 1987b) and categories derived from the literature (Woolsey, 1987a). Second, relevant items were selected from the Affiliation Questionnaire (Toews, 1973) and rewritten to fit the format of the Bonds Scale. The resulting 192-item questionnaire was completed by 1,379 men and women from student (university and community college) and community (non-credit community adult education classes) samples. In terms of age, marital status, and occupational groupings, the characteristics of the four samples - student women (n = 708), student men (n = 363), community women (n = 188), and community men (n = 120) were similar to those described for the Close Social Relationships Scale. As with the Close Social Relationships Scale, unless noted otherwise, the Bonds scale development described in this section has been extracted from the, as yet, unpublished data analyses of Woolsey et al. (1988a). Each female respondent was requested to answer each item on the 192-item Bonds Scale according to how strongly she believed it to be true or not true of her general attitudes and feelings about women. Male respondents were requested to respond according to their general attitudes  103 and feelings about men. The 6-point scale for responses ranged from 1 (definitely not true) to 6 (definitely true). At this point, the Items on the men's and women's scales were identical except for the appropriate interchanging of the words "women" and "men" as described above. Principal component analyses were performed separately on the data gathered from each of the four different groups: student women, student men, community women, and community men. Five factors emerged from these initial analyses: (a) Cross-Gender Preference, (b) Same-Gender Preference, (c) Valuing, (d) Solidarity, and (e) Identification. At this time, Identification appeared to be a viable factor only for women (both student and community samples) and Solidarity appeared to have some items that were common to women and men in all four samples, some items that worked only for women (student and community), and some that worked only for men (student and community). At this point, items were retained or dropped according to the following decision rules: (a) the Identification items were retained only for women (this decision was confirmed by conceptual analysis); (b) each of the items was retained only if it had a factor loading of .30 or more on one of the four factors (for men) or five factors (for women); (c) to be retained, an item had to be factorially simple (i.e., it had a loading of .30 or more on one factor only). Several rounds of unweighted least squares factor analysis (four factors for men, five for women), using Harris-Kaiser oblique transformation, were performed on the retained items from the original data sets. Each round of factor analysis was followed by the retention or deletion of items according to the above decision rules. Eventually, the Solidarity items that appeared to work only for one gender were dropped from the analyses for the other gender (again, this decision was confirmed by conceptual analysis). The final round of Harris-Kaiser transformed unweighted least squares factor analysis was performed with 60 items and four factors for men, and 65 items and five factors for women. The same four factors (for men) and five factors (for women) remained in this last round of analysis. The final 58 (for men) and 62 (for women) items retained from this round were factorially simple and had primary-factor pattern coefficients greater than or equal to .30 for all four samples, with the exception of Identification (high loadings for women only) and some Solidarity items (for women or  104 men only). Oblique primary-factor intercorrelations obtained in this final round of factor analysis remained acceptably low, ranging in magnitude from .00 to .30. (Factors and correlations among them described here are for the best [oblique] solutions obtained in the final round of factor analysis.) The 58 (for men) and 62 (for women) retained items form the core of the Bonds Scale. While scale development is continuing, the psychometric properties of this form of the scale were strong enough to justify its use in the current study. However, based on further analyses conducted on the scale in preparation for the present study, some changes were made to the scale for the purpose of the current study. The internal consistency for the Solidarity Scale was very low for all four samples (.46 and .40 for the 13-item scale for student and community men, respectively; .38 and .23 for the 11 -item scale for student and community women, respectively). Given that the conceptual strength of the factor was in the women-only items regarding jealousy, these 4 items were retained (and scored so that a high score indicated a high level of jealousy) and the rest were dropped. This resulted in the alpha rising markedly to .79 for both the women's samples. No equivalent men-only core of items could be found. Therefore, the Solidarity Scale was totally dropped for men, leaving three factors or subscales for men. All of the Solidarity items, except the 4 women-only jealousy items, were deleted for women and unweighted least squares factor analyses (using Harris-Kaiser oblique transformations) were performed on the reduced item sets for both female samples. Five factors remained in the same structure as in previous analyses. Given the strong factorially simple structure and the consistently high level of internal consistency for all five subscales for women, these five scales were used in the current study. In summary, the three factors common to women and men are Cross-Gender Preference, Same-Gender Preference, and Valuing. The two women-only factors are Identification and Jealousy. For the purpose of the current study, the 2 least strong Items of the Cross-Gender Preference Scale were dropped, bringing the number of items on this scale down to 15. The 3 least strong SameGender Preference Items were dropped, bringing the number of items in that scale down to 15 as well. All 10 Valuing items were retained, as were all 6 Identification items and all 4 Jealousy items.  105 Interscale correlations for the resulting 50-item Bonds Between Women Scale and 40-item Bonds Between Men Scale were acceptably low, ranging in magnitude from .00 to .54 for the student and community samples of women and men. The final version of the women's scale used in the current study contained 60 items due to the addition of 4 new Identification items and 6 new Jealousy items. The additional 4 Identification items and 6 Jealousy items were drawn from a revised version of the Bonds for which separate data have been collected but not analyzed (Woolsey et al., 1988a). The 60-item Bonds Between Women Scale used in the current study is found in Part Two of the sample questionnaire (see Appendix A). Item numbers for each of the five subscales are (new items in bold print and reverse-scored items underlined): Valuing (1,5, 10,15, 20, 25, 30, 35, 39, 44); Cross-Gender Preference (2, 6,11,16, 21, 26, 31, 36, 40, 45, 49, 51, 53, 56, 59); Same-Gender Preference (3, 8, 13, 18, 23, 28, 33, 37, 42, 47, 50, 52, 55, 58, 60); Identification (4, 9, 14, 19,24,29, 34,38, 43, 48); and Jealousy (7,12,17, 22,27,32,41, 46,54, 57). The 40-item Bonds Between Men Scale used in the current study (see Appendix A) contains the following subscales and items (reverse-scored items underlined): Valuing (1,4, 7, 10, 13, J6.19, 22, 25, 28); Cross-Gender Preference (2, 5, 8,11,14,17, 20, 23, 26, 29, 31, 33, 35, 37, 39); and Same-Gender Preference (3, 6, 9, 12,15, 18, 21, 24, 27, 30, 32, 34, 36, 38, 40).  Scoring. Each item on the Bonds Scale receives a score from 1 (definitely not true) to 6 (definitely true). A separate score is derived for each of the five (for women) and 3 (for men) subscales by reversing the value of reverse-scored items and then summing the score for all items contained in the subscale.  Reliability. Internal consistency reliability coefficients (computed from the student and community data) for each of the subscales of the 50-item Bonds Between Women Scale and the 40item Bonds Between Men Scale ranged from .77 to .90. Test-retest reliability has not been investigated.  106  Validity. Evidence of convergent validity for four subscales - Cross-Gender Preference (29 items), Same-Gender Preference (29 items), Valuing (10 items), and Identification (7 items) -- of an earlier 75-item version of the Bonds Between Women Scale was provided by Sigal's (1987) study of the homosocial bonds, liberated behaviour, and self-esteem of 112 single, childless women. As hypothesized, nontraditional, feminist women (i.e., women with high scores on the Liberated Behavior Questionnaire [Ghaffaradli-Doty & Carlson, 1979]) had significantly higher mean scores on Same-Gender Preference, Valuing, and Identification and significantly lower mean scores on CrossGender Preference than did traditional women. Largely unsupported were predictions that women with high scores on Same-Gender Preference, Valuing, and Identification and low scores on CrossGender Preference would obtain significantly higher self-esteem and mastery scores than would women with low scores on Same-Gender Preference, Valuing, Identification and high scores on Cross-Gender Preference. Convergent validity also was investigated in the present study by testing the hypotheses that scores on the Bonds subscales would be predictive of (a) the sizes of same- and cross-gender networks, and (b) mentoring behaviours and relationship provision scores. These results are reported in Chapter IV (Results). Construct validity has been assessed, in part, by the evidence of convergent validity cited above and by repeated factor analyses (Woolsey et al., 1988a) described earlier which resulted in the same factor structure of conceptually meaningful factors. This process was continued with factor analyses of the data collected for the current study (these results are reported in Appendix C [Preliminary Analyses]).  3. Supportive (Mentoring) Functions and Behaviours Development of the Scale. Schroeder (1988/1989) wrote the items for the supportive functions and behaviours measures based on an extensive review of the literature on mentoring roles, functions, and behaviours. Originally, six items were written to represent each of nine mentor  107 functions. Six of the functions were classified as career functions which "enhance learning the ropes and preparing for advancement in an organization" (Kram, 1985, p. 22). These functions were Sponsoring, Exposure and Visibility, Teaching the Job, Teaching the Informal System, Protection by Prevention, and Protection by Absorption. The remaining three functions were classified as psychosocial functions which "enhance a sense of competence, clarity of identity, and effectiveness in a professional role" (Kram, 1985, p. 22). These functions were Role Modeling, Encouragement, and Personal Counselling. To verify that the items written were representative of the functions for which they were written, five naive judges were asked to sort the items according to which function they believed each item to represent. Three of the five judges correctly sorted all 54 items; the remaining two judges incorrectly sorted one item each. Because of the high agreement between the judges, all 54 items were included in a pilot study of the scale. The results of the pilot study (18 respondents from the Sales and Marketing Diploma Program at UBC) led to several changes in the original version of the questionnaire. The low response rate (21.9%) was attributed, in part, to questionnaire design (too many items rated for too many people). Therefore, the questionnaire was shortened in the following ways. First, the number of items for each function was reduced from six to five by deleting the item on each scale which least affected the internal consistency of the function. Second, Protection by Absorption and Protection by Prevention were collapsed into one 5-item function (Protection) because of the high correlations between the two functions. Third, because the perceived effect of each activity was found to be highly correlated with the extent to which ft was provided, it was decided to assess extent only. That is, each item was to receive one rating from 0 (never) to 4 (frequently) according to the exfenf to which each person identified engaged in that behaviour on behalf of the respondent. Finally, changes were made to the number of people to be rated by each respondent and to the manner in which the initials of these people were elicited. The revised questionnaire asked respondents to list the initials of people in their organization who were currently providing them or who had provided them with the following eight mentor functions (in this order): Sponsoring, Role Modeling, Teaching the Informal System, Personal Counselling, Teaching the Job,  108 Exposure and Visibility, Encouragement, and Protection. The format and questions (items 1 to 8) used in this identification process are located under Supportive Functions in Part One of the questionnaire used in the present study (see Appendix A). In the current study, the procedure for identifying providers of mentoring behaviours was nearly identical to that used by Schroeder (1988/1989), the main difference being the maximum number of people respondents were asked to list (five in the current study, three in Schroeder's). In his pilot study, Schroeder found that people listed an average of three people for each function. When people in the pilot study were asked to list up to five providers and to rate each person regardless of the number of functions provided, persons listed as providing less than three functions were subsequently rated as providing very low levels of mentoring behaviours overall. Thus, information about these people provided very little additional information and responding to individual items in regard to these people may have caused the respondents to become disillusioned with the questionnaire. In his revised questionnaire, Schroeder requested respondents to answer each of the 40 mentoring activities for up to three of the people listed most often as providers of mentor functions. The person listed most frequently was rated regardless of the number of functions he or she had been identified as providing. A second or third person was rated only if he or she provided at least three mentoring functions. The provision of mentoring behaviours by networks of individuals (i.e., same-gender, crossgender) as opposed to single individuals, was the focus of the present study. To prevent the loss of information about the size and gender composition of the networks as a whole, respondents were asked to list up to five persons for each mentoring function. Then, rather than rating individuals separately on each of the 40 mentoring activities, each network (up to 10 same- or cross-gender persons listed most frequently) was rated as a whole on each of the 40 mentoring activities. The format and instructions for identifying the networks to be rated are located (a) under the title Network Table (which follows the Supportive Functions section of the questionnaire), and (b) in the introductory section under the title Supportive Behaviours in Part One of the questionnaire (see  109 Appendix A). This procedure was a modification of procedures used for similar purposes by Fischer (1982), Schroeder (1988/1989), and Woolsey et al. (1988b). Schroeder's (1988/1989) revised questionnaire was completed by 442 management, supervisory, professional, and technical employees in five different organizations (in energy, food wholesaling, petroleum, communications, and mining industries in British Columbia). The overall response rate for distributed questionnaires was 70.8% (442 of 624). Three hundred and ninetyseven of the 442 questionnaires returned were useable. The average age of the sample of respondents (n = 397) was 37.99 years (SD = 7.06) and the average length of organizational tenure was 10.31 years (SD = 4.13). One hundred eleven (28.0%) of the respondents were women; 286 (72.0%) were men. Depending on the number of people respondents listed as providing three or more mentoring functions, each respondent rated between one and three people on each of the 40 mentoring behaviours listed under Supportive Behaviours in Part Two of the questionnaire used in the current study (see Appendix A). The mentoring behaviour scale items corresponding to the eight mentoring functions are as follows: Sponsoring (1,9,17,25, 33); Role Modeling (2,10,18, 26, 34); Teaching the Job (3,11,19, 27, 35); Personal Counselling (4,12, 20, 28, 36); Exposure and Visibility (5,13, 21, 29, 37); Encouragement (6,14, 22, 30, 38); Protection (7,15, 23, 31, 39); and Teaching the Informal System (8,16, 24, 32, 40). The results of principal component analyses performed on the ratings obtained for the eight mentoring behaviour scales suggested that "the eight supportive [mentoring] behaviour scales loaded on one general component that accounted for over 5 0 % of the variance" (Schroeder, 1988/1989, p. 108). When principal component analyses were performed on the individual items (as opposed to the subscales as described above), six to eight components emerged. The items loading on each component were consistent with the item content of the original eight mentoring behaviour scales which were conceptually derived (K.G. Schroeder, personal communication, 1988). However, "the high correlations among the supportive [mentoring] behaviour scales... indicate that respondents who received relatively high levels of behaviours associated with one function also  110 received relatively high levels of behaviours associated with the other functions" (Schroeder, 1988/1989, p. 154). The pattern of results noted above lent little empirical support for a separation of mentoring behaviours into career functions and psychosocial functions. Rather, the overall level of mentoring behaviours provided was indicated as the most meaningful unit of analysis. Schroeder (1988/1989) called this one general supportive behaviour component the Total Supportive Behaviour Score (TSBS). The TSBS was the unit of analysis that was adopted in the current study. The psychosocial aspect of organizational relationships was measured by the Close Social Relationships Scale (Woolsey et al., 1988b).  Scoring. Each of the 40 supportive (mentoring) behaviour items is rated according to the extent to which each source of mentoring behaviours engages in that activity on behalf of the respondent. The 5-point rating scale ranges from 0 (never) to 4 (frequently). A separate score may be derived for each mentoring behaviour scale by summing the responses to each of the five items in the scale. The Total Supportive Behaviour Score (TSBS) is derived by adding the scores obtained for each of the eight mentoring behaviour scales. All items are positively worded.  Reliability. Based on the results obtained in Schroeder's (1988/1989) analysis of the data from 397 respondents, the individual supportive behaviour scales and the global supportive behaviour scale were judged to have adequate internal consistency. Alpha coefficients ranged from .78 to .92 for each of the eight supportive behaviour scales (alphas derived separately for each of the three sources of mentoring behaviours). The alphas for the TSBS were .96, .96, and .95 for the first, second, and third sources of mentoring behaviours, respectively. Test-retest reliability has not been assessed.  Validity. Convergent validity of the supportive behaviours scales was assessed by comparing the ratings for people who were viewed as mentors with ratings of people who were not  Ill viewed as mentors (according to two different definitions of mentors used in previous research). For both definitions, persons who were viewed as mentors received significantly higher mean scores on all eight supportive behaviour scales and on the TSBS than persons who were not viewed as mentors. In addition, persons who were viewed as mentors according to only one mentor definition received significantly lower mean global ratings (TSBS) than persons who were viewed as mentors according to both definitions and significantly higher mean global ratings (TSBS) than persons who were not viewed as mentors according to both definitions (Schroeder, 1988/1989). The 397 respondents in Schroeder's (1988/1989) study were similar in age, gender, occupation, and job tenure to the respondent population proposed for the current study. Therefore, use of Schroeder's supportive functions and behaviours measures with the respondents in the present study was judged to be valid.  4. Satisfaction With Progression Satisfaction with progression within the organization was assessed using a 5-item scale developed by Schroeder (1988/1989). This variable was conceptualized as "employee satisfaction with changes in salary, positions, and professional development over the course of organizational tenure" (Schroeder, 1988/1989, p. 76). The instructions and items (numbers 16 to 20) for this scale are listed in Part One of the questionnaire (see Appendix A). The items are scored on a 5-point scale ranging from 1 (very unsatisfied) to 5 (very satisfied). All items are positively worded, so a high score indicates a high degree of satisfaction with progression. Schroeder (1988/1989) found the internal consistency of this scale to be .84 in his pilot study and .81 in his main study.  5. Average Annual Salary Increase This variable was assessed by averaging the respondent's yearly increase in salary over the length of employment in the current organization. The three items used to gather information for this variable were placed in the Demographic Measures section of Part One of the questionnaire (see  112 Appendix A, Demographic Measures: item numbers 6,10,11). Schroeder (1988/1989) used the same items and method of assessing average annual salary increase in his study.  6. Average Number of Promotions per Year This variable was assessed by dividing the number of promotions obtained since the respondent first joined his or her present organization by the number of years of employment with that organization. The two items used to gather information for this variable (item numbers 6 and 9) are listed under Demographic Measures in Appendix A. The same items and method of assessment were used by Schroeder (1988/1989).  7. Skill Development Three types of skill development were assessed: job, interpersonal, and conceptual. The conceptual basis for these scales was provided by Katz (1955) and Reber and Van Gilder (1982) who identified three types of skills (technical, human, and conceptual) that are important for successful administration. Technical (job) skill refers to "a person's proficiency in, and understanding of, the specific techniques, processes, methods, and procedures required in carrying out a particular job" (Reber & Van Gilder, 1982, p. 12). Human (interpersonal) skill is "the ability to interact effectively with people" (Reber & Van Gilder, 1982, p. 12). Finally, conceptual skill is defined as a manager's ability to "visualize the organization as an integrated whole, recognizing how a change in any one part affects all the other parts. It's the ability to see the big picture'" (Reber & Van Gilder, 1982, p. 12). The individual scales developed to measure each of these three skills are described separately below.  Job Skill Development. Job skill development was assessed using a list of 25 separate job skills identified by Pinder (1982) and Pinder and Schroeder (1987). The degree to which the respondent has developed each of the 25 skills since joining his or her present organization is rated on a 5-point scale ranging from 0 (not at all) to 4 (a great deal). The 25 skills are listed in Appendix A  113 under Job Skill Skill Development; the instructions and response format for the measure as listed in Appendix A were developed by Schroeder (1988/1989). Pinder's (1982) attempts to form subscales from the items in the job skill development scale were unsuccessful. The average inter-item correlations in two studies were .48 (Pinder, 1982) and .31 (Schroeder, 1988/1989). The internal consistency of the scale was .96 in Pinder's (1982) study, .92 in Schroeder's (1988/1989) pilot study, and .93 in Schroeder "s (1988/1989) main study. Only 13 (3.7%) of the 356 respondents in Pinder's (1982) investigation stated that they had developed skills other than those listed on the job skill development scale, indicating that the scale is a relatively complete list of the skills required in daily work.  Interpersonal Skill Development. This 5-item scale was developed by Schroeder (1988/1989). He wrote one item to assess each of the five qualities or attitudes identified by Reber and Van Gilder (1982) as important for the development of a manager's effectiveness in interpersonal relationships. These qualities or attitudes are: empathy, acceptance of individual differences, self-awareness, perceptual awareness, and an employee orientation. These items (numbers 3, 6, 9,12, and 15) are listed in Appendix A under Acceptance by Co-workers, Conceptual Skill Development, Interpersonal Skill Development. Each of these items is scored on a 7-point scale ranging from 1 (very false) to 7 (very true). The internal consistency of the scale was .77 in Schroeder's (1988/1989) study.  Conceptual Skill Development. This 5-item scale was developed by Pinder (1982). The items refer to the ability to recognize how the organization operates as a whole, that is, the ability to understand "the bigger picture," as discussed by Reber and Van Gilder (1982). Each item is scored on a 7-point scale ranging from 1 (very false) to 7 (very true). The items (numbers 2, 5, 8,11, and 14) are listed in Appendix A under Acceptance by Co-workers, Conceptual Skill Development, Interpersonal Skill Development. The internal consistency of the conceptual skill development scale  114 was .86 in Pinder's (1982) study, .82 in Schroeder's (1988/1989) pilot study, and .82 in Schroeder's (1988/1989) main study.  8. Job Satisfaction Job satisfaction was assessed using the 5-item general satisfaction scale of the Job Diagnostic Survey (Hackman & Oldham, 1974). The Job Diagnostic Survey has been used widely in research and job change projects (Hackman & Oldham, 1980; Oldham, Hackman, & Stepina, 1979). The general satisfaction scale is a measure of personal affective reactions to a job in general. The items for this scale (numbers 30, 31, 32, 33, and 34) are listed under Job Satisfaction in Appendix A. Three of the items (numbers 30, 31, and 32) assess general satisfaction directly by asking respondents to rate their own feelings about their job. Two items (numbers 33 and 34) assess general satisfaction indirectly by asking respondents to rate the general reactions of ofner people (in jobs similar to the respondent's in the organization) to their jobs. Each item is rated on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). Item numbers 31 and 34 are reversed prior to scoring (i.e., 7=1, 6=2, 5=3, 4=4, 3=5, 2 = 6,1 =7). The internal consistency of the general satisfaction scale has been found to be .76 (Hackman & Oldham, 1974) and .77 (Oldham et al., 1979) in two different studies.  9. Organizational Commitment The 9-item short form of the Organizational Commitment Questionnaire (OCQ) (Mowday, Steers, & Porter, 1979) was used to assess the commitment of the individual to the organization. Organizational commitment is defined as the "relative strength of an individual's identification with and involvement in a particular organization" (Mowday et al., 1979, p. 226). It is "a general affective response to the organization as a whole  [it] emphasizes attachment to the employing  organization, including its goals and values" (Mowday et al., 1979, p. 226). As such, organizational commitment differs from job satisfaction which emphasizes affective reactions to the specific job environment where an employee carries out his or her work (Mowday et al., 1979). The items for the  115 OCQ (numbers 21 to 29) are listed under Organizational Commitment in Appendix A. The items are scored on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). As all of the items are positively worded, a high score indicates a high level of commitment to the organization. Mowday et al. (1979) reported a range of internal consistency from .84 to .90 when the OCQ was used with diverse samples (e.g., scientists and engineers, university employees, bank employees). Schroeder (1988/1989) found the internal consistency to be .93 in his pilot study and .92 in his main study. Convergent validity was assessed by correlating scores on the OCQ to scores on a measure assessing employees' intentions to leave the organization. The relatively high negative correlations (i.e., ranging from -.31 to -.63 in diverse samples) gave evidence of convergent validity (Mowday etal., 1979). Evidence of discriminant validity was less conclusive. While the percentage of variance shared by organizational commitment and measures such as career satisfaction, job satisfaction, and job involvement was generally under 2 5 % and not over 50%, the correlations between the scales were higher than would have been desirable for conclusive evidence of discriminant validity.  10. Acceptance by Co-workers  Acceptance by co-workers was assessed using a 5-item scale developed by Schroeder (1988/1989). The scale was designed to measure the degree to which an employee's co-workers accept the employee's beliefs, opinions, and judgement. The content of the scale was suggested by another 5-item co-worker support scale used by Pinder and Schroeder (1987). The items for this scale (numbers 1, 4, 7,10, and 13) can be found under Acceptance by Co-workers, Conceptual Skill Development, Interpersonal Skill Development in Appendix A. Each item is rated on a 7-point scale ranging from 1 (very false) to 7 (very true). Item numbers 4 and 13 are reverse scored. Schroeder (1988/1989) found the internal consistency of this scale to be .83.  116 11. Loneliness at Work Loneliness at work was assessed using an adapted version of the 20-item revised UCLA Loneliness Scale (Russell, Peplau, & Cutrona, 1980). Loneliness "reflects an individual's subjective perception of deficiencies in his or her network of social relationships" (Russell, Cutrona, Rose, & Yurko, 1984, p. 1313). Presently, the UCLA Loneliness Scale is the most widely used measure of loneliness (Russell, 1982). Scale development began with sampling 25 items from Sisenwein's (1964/1965) loneliness scale. Respondents in two samples of young adults at UCLA (clinical sample and student sample) responded to each of the 25 items by rating on a 4-point scale how frequently they felt the way described in the item. In addition, the participants rated how lonely they were compared to other persons. They also described their present affective state by rating the intensity of feelings such as "empty," "restless," "bored," and "depressed." The final loneliness scale consisted of 20 items selected on the basis of high item-total correlations (retained items had correlations above .50). Internal consistency for this scale was .96. This version was eventually revised for three reasons: (a) because all items on the scale were worded in the same (lonely) direction, there was concern about response bias; (b) because of the social stigma associated with loneliness, it was thought that respondents might distort their responses to appear less lonely; and (c) the magnitude of the correlations between the loneliness scale and other measures such as self-esteem and depression indicated a problem with discriminant validity. The revised UCLA Loneliness Scale was developed to address these issues. The revised UCLA Loneliness Scale has 10 positively scored items and 10 negatively scored items. Two studies using the revised scale have shown evidence for validity (Russell et al., 1980). In the first study concurrent validity was established by finding significant positive correlations between loneliness scores and measures of depression and anxiety. Loneliness scores were not significantly related to scores on measures conceptually unrelated to loneliness (e.g., feeling creative, embarrassed, or surprised). In the second study, evidence of concurrent validity was obtained by examining the relationship between loneliness scores and measures of social activities and  117 relationships. The discriminant validity of the revised UCLA Loneliness Scale was indicated by findings that the loneliness scores were not confounded by social desirability. Further evidence of discriminant validity was indicated by findings that revised UCLA loneliness scores were more highly correlated with other measures of loneliness than with measures of personality variables and mood. Internal consistency was found to be .94 in both studies reported by Russell et al. (1980). The 20 items for the revised UCLA Loneliness Scale can be found under Loneliness at Work in Appendix A. The words "at work" have been inserted into each item to ensure that respondents respond to the items with their work setting in mind. Each item is scored on a 4-point scale ranging from 1 (never) to 4 (often). Item numbers 1, 4, 5, 6, 9,10,15,16,19, and 20 should be reversed (i.e., 1 =4, 2=3, 3=2, 4 = 1) prior to summing the scores for all 20 items to obtain a total loneliness score.  12. Demographic Measures The items in the Demographic Measures section of the questionnaire (see Appendix A) were used to gather information about each respondent's age, gender, marital status, education, ethnic designation, length of organizational tenure, number of promotions, current salary, initial salary, and current job category.  C. DESIGN AND  STATISTICAL ANALYSIS  The design and methods of statistical analysis employed in the current study are outlined in this section. The variables, measures, design, and analysis for each of the five research questions are summarized in Figures 2 to 6. The independent variables and the measures utilized in gathering the independent variable data are listed across the top portion of each figure; the dependent variables and the measures associated with them are listed across the middle section of each figure. Finally, the design and methods of statistical analysis are summarized at the bottom of each figure. The large number of statistical analyses conducted to test the hypotheses for this study greatly increased the risk of Type I error. To counteract this risk, the alpha for tests of significance  Independent Variables Measure: * Bonds Between Women/ Bonds Between Men Scale  Cross-Gender Preference  Same-Gender Preference  Identification (women only)  Jealousy (women only)  Dependent Variables Measure: * Network Table completed in preparation for administration of Schroeder's (1988/1989) supportive (mentoring) behaviours measure  Design and Analysis:  Size of same-gender network  Size of cross-gender network  Standard multiple regression analysis and one-tailed tests of significance of the bivariate correlations between independent and dependent variables. Each dependent variable was regressed separately onto homosocial attitudes (three scales for men, five scales for women). Separate analyses were conducted for women and men, making a total of four regression analyses. Alpha was set at .0125.  Figure 2. Variables, measures, design, and analysis for Research Question 1: What is the relationship between an individual's homosocial attitudes and the size of his or her same- and cross-gender networks?  Cross-Gender Preference  Independent Variables  Same-Gender Preference  Valuing  Measure: * Bonds Between Women/ Bonds Between Men Scale  Identification (women only)  Jealousy (women only)  Dependent Variables Measures: * Schroeder's (1988/1989) supportive (mentoring) behaviours measure * Close Social Relationships Scale  Design and Analysis:  Total Supportive (Mentoring) Behaviour Score  Standard multiple regression analysis and one-tailed tests of significance of the bivariate correlations between independent and dependent variables. Each dependent variable was regressed separately onto homosocial attitudes (three scales for men, five scales for women). Separate analyses were done for men and women and for the provisions of same- and cross-gender networks, bringing the total number of regression analyses for this question to 20. Alpha was set at .0025.  Figure 3. Variables, measures, design, and analysis for Research Question 2: What is the relationship between an individual's homosocial attitudes and the levels of supportive (mentoring) behaviours and relationship provisions he or she receives from same- and cross-gender networks?  Independent Variables  Gender composition of network (same-gender, cross-gender)  Dependent Variables Measures: * Schroeder's (1988/1989) supportive (mentoring) behaviours measure * Close Social Relationships Scale  Total Supportive (Mentoring) Behaviour Score  Relationship provisions: • Intimacy • Similarity • Defiance of Convention • Respect for Differences  Design and Analysis:  1.  2 x 2 ANOVA and simple main effects analysis for the significant interaction (for the supportive [mentoring] behaviours portion of the question).  2.  2 x 2 MANOVA followed up with four ANOVAs and simple main effects analysis for one significant interaction (for the relationship provisions portion of the question).  Alpha was set at .025.  Figure 4. Variables, measures, design, and analysis for Research Question 3: Are there gender differences in the levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks?  Independent Variables Measures: * Network Table and Schroeder's (1988/1989) supportive (mentoring) behaviours measure * Bonds Between Women/ Bonds Between Men Scale  Level of mentoring received from same-gender networks (Low, High)  Same-Gender Preference Composite (Low, High)  Level of mentoring received from cross-gender networks (Low, High)  Cross-Gender Preference Composite (Low, High)  Dependent Variables Measures: * General satisfaction scale (from the Job Diagnostic Survey) * Organizational Commitment Questionnaire * Acceptance by Co-workers Scale * Revised UCLA Loneliness Scale * Satisfaction With Progression Scale * Average Annual Salary Increase * Average Number of Promotions per Year * Job Skill Development * Interpersonal Skill Development  Affective Outcomes  Advancement Outcomes  Skill Development Outcomes  Affective Outcomes  Advancement Outcomes  Skill Development Outcomes  * Conceptual Skill Development Design and Analysis:  Four 2 x 2 MANOVAs, then each significant MANOVA was followed up with three ANOVAs. Scores for the independent variables were dichotomized at the median for each variable. Separate analyses were conducted for Level of Mentoring Received From Same-Gender Networks x Same-Gender Preference Composite and Level of Mentoring Received from Cross-Gender Networks x Cross-Gender Preference Composite. Factor scores for each of the three dependent variables were derived through maximum likelihood factor analysis of the conflation matrix for the 10 career- and job-related outcome variables listed above as measures of the dependent variables. Separate analyses were conducted for women and men. Alpha was set at .0125.  Figure 5. Variables, measures, design, and analysis for Research Question 4: How do homosocial attitudes in interaction with levels of supportive (mentoring) behaviours received from same- and cross-gender networks relate to career- and job-related outcomes?  M  •H  Independent Variables Measures: * Schroeder's (1988/1989) supportive (mentoring) behaviours measure * Close Social Relationships Scale  Total Supportive (Mentoring) Behaviour Score  Intimacy  Similarity  Defiance  Respect  of  for  Convention  Differences  Dependent Variables Measures: * General satisfaction scale (from the Job Diagnostic Survey) * Organizational Commitment Questionnaire * Acceptance by Co-workers Scale * Revised UCLA Loneliness Scale * Satisfaction With Progression Scale * Average Annual Salary Increase * Average Number of Promotions per Year * Job Skill Development * Interpersonal Skill Development * Conceptual Skill Development  Design and Analysis:  Affective  Advancement  Outcomes  Outcomes  Skill Development Outcomes  Standard multiple regression analysis and one-tailed tests of significance of the bivariate correlations between independent and dependent variables, followed up with t-tests to test the differences between the bivariate correlations that were different in the hypothesized direction. Each dependent variable was regressed separately onto the five independent variables. Separate analyses were conducted for women and men and for the provisions of same- and cross-gender relationships, bringing the total number of regression analyses for this question to 12. Alpha was set at .0042.  Figure 6. Variables, measures, design, and analysis for Research Question 5: How do levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks relate to career- and job-related outcomes?  123 was apportioned using the Bonferroni inequality procedure (Dunn, 1961; Marascuilo & Levin, 1983) to maintain an error rate of 5 % within the first set of analyses conducted for the hypotheses associated with each research question. The alpha for the tests of significance associated with each research question is given at the end of the Design and Analysis portion of each figure (see Figures 2 to 6). Because scale development was in progress for several key variables investigated in the present study (career- and job-related outcomes, homosocial attitudes, relationship provisions), it was necessary to conduct factor analyses, utilizing data gathered in the current study, to explore the factor structures of these variables in preparation for the hypothesis-testing analyses outlined in Figures 2 to 6. The results of these factor analyses are presented in detail in Appendix C (Preliminary Analyses), and are described briefly below in order to explain some of the variables presented in Figures 2 to 6.  Career- and job-related outcomes.  Separate scores were obtained on each of the following  10 measures: Satisfaction With Progression, Average Annual Salary Increase, Average Number of Promotions Per Year, Job Skill Development, Interpersonal Skill Development, Conceptual Skill Development, Job Satisfaction (General Satisfaction Scale of the JDS), Organizational Commitment, Acceptance by Co-workers, and the revised UCLA Loneliness Scale. Factor analyses performed on the correlation matrices for these scales reduced the original 10 variables to three factors, for both women and men, replicating the basic three-factor structure identified by Schroeder (1988/1989) in his principal component analysis of 11 similar variables. Factor scores obtained on each of the three factors ~ Affective Outcomes, Advancement Outcomes, and Skill Development Outcomes - were used as the measures of the dependent variables for Research Questions 4 and 5 (see Figures 5 and 6).  Homosocial attitudes. The Bonds Between Women/Bonds Between Men Scale data obtained from women and men in the current study were factor analyzed separately because of  124 differences in the item content of the scales. Factor analyses of the items replicated the basic fivefactor simple structure (for women) and three-factor simple structure (for men) obtained by Woolsey et al. (1988a). The three factors common to women and men are Cross-Gender Preference, SameGender Preference, and Valuing. The two women-only factors are Identification and Jealousy. Scale scores for each of the five factors (for women) and three factors (for men) were used as the measures of the independent variables for Research Questions 1 and 2 (see Figures 2 and 3). Subsequent factor analysis performed on the correlation matrix for the five Bonds scales for women, and principal component analysis performed on the three scales for men, produced two composite factors which were conceptually comparable for women and men. Factor scores on these two composite factors -- Same-Gender Preference Composite and Cross-Gender Preference Composite -- were used as the measures of two of the independent variables for Research Question 4 (see Figure 5).  Relationship provisions. The Close Social Relationships Scale was completed by women and men with respect to the provisions of same- and cross-gender networks, thereby producing four different CSRS data sets to be analyzed: (a) women's same-gender, (b) women's cross-gender, (c) men's same-gender, and (d) men's cross-gender. For each of the four different data sets, factor analysis was performed on the items. Only four of the six factors found in previous research (Woolsey et al., 1988b) were replicated; in the current study, items were dropped after each round of factor analysis if they did not have a factor loading greater than or equal to .30 on the same conceptually meaningful factor for all four administration conditions. Scale scores on each of the four replicated factors -- Intimacy, Similarity, Defiance of Convention, and Respect of Differences -were used as the measures of four of the dependent variables in Research Questions 2 and 3 (see Figures 3 and 4) and as the measures of four of the independent variables in Research Question 5 (see Figure 6). The fifth and sixth factors not supported by these factor analyses - Emotional Support and Lifetime Attachment - were dropped from all further analyses for the present study.  125 CHAPTER IV. RESULTS  The research findings are presented in this chapter. The chapter begins with a discussion of the characteristics of the respondent sample. Following that, the hypotheses associated with each of the five research questions introduced in the preceding chapters are evaluated. Preliminary analyses conducted in preparation for the hypothesis-testing analyses are presented in Appendix C. The preliminary analyses include procedures for handling missing data and factor analyses performed to explore the factor structures of key variables used in the study (career- and job-related outcomes, homosocial attitudes, and relationship provisions). In addition, because the psychometric properties of many of the measures employed in this study have not been reported elsewhere, the scale means, standard deviations, internal consistency reliability coefficients, and so forth of all measures used in the current study also are presented in Appendix C.  A. CHARACTERISTICS OF THE RESPONDENT SAMPLE  1. Number of Respondents Information about the number of distributed, returned, excluded, and retained questionnaires is contained in Table B-1 (see Appendix B). The results in Table B-1 are presented separately for women and men, and for each of the seven participating organizations. The organizations have been broadly classified according to the industrial sector from which they were drawn (public sector, private sector, Crown corporation). The overall response rate for women was 39.0% (283 of 725 questionnaires were returned); the women's response rate for returned questionnaires within each organization ranged from 32.3% (post-secondary education) to 92.3% (utilities). Of the 283 questionnaires returned by women, 26 were excluded from analyses: 7 because of a large number of missing responses, 17 because the respondents had been employed in their current organization for less than 6 months or more than 17  126 years, and 2 because the respondents failed to identify the name of their organization. The women's response rate for useable questionnaires, then, was 35.5% (257 of 725). The overall response rate, and the response rate within most of the individual organizations, was lower for men than for women. Overall, the men's response rate for returned questionnaires was 29.6% (216 of 731); the rate within each organization ranged from 23.1% (post-secondary education) to 100.0% (equipment sales and service). Nineteen of the 216 questionnaires returned by men could not be used in the analyses: 5 because of a large number of missing responses, 13 because the respondents had less than 6 months or more than 17 years tenure with their current organization, and 1 because the respondent omitted the name of his organization. The men's response rate for useable questionnaires, then, was 27.0% (197 of 731). One of the criteria for selecting potential respondents for the study was that respondents have no more than 15 years of employment with their current organization. The 9 women and 1 man who stated that they had been employed in their current organization for 16 or 17 years were included in analyses because they were so close to the cutoff of 15 years and because their responses were needed to increase the n, particularly for the factor analyses. Ten women and 2 men with more than 17 years organizational tenure were excluded from analyses because their lengths of organizational tenure were well beyond the 15-year cutoff point. The 7 women and 11 men with less than 6 months organizational tenure were excluded from analyses because their frequency distributions for variables such as supportive (mentoring) behaviours, relationship provisions, and average number of promotions per year showed a very different pattern of responses than those for respondents with 6 months or more organizational tenure. The different patterns of responses were attributed to a lack of time (i.e., less than 6 months) to form relationships, to be promoted, and so forth. In summary, the overall response rate for returned questionnaires was 34.3% (499 of 1,456); considered separately by gender, the overall response rate was 39.0% for women and 29.6% for men. A total of 45 questionnaires (3.1% of those returned) could not be used in analyses because of incomplete responses, because the respondents had less than 6 months or more than 17 years  127  organizational tenure, or because respondents had not specified the name of their organization. The response rate for useable questionnaires was 31.2% (454 of 1,456) overall, or 35.5% for women and 27.0% for men. The response rates for useable questionnaires varied greatly by organization and industrial sector. The lowest response rates were obtained from the two post-secondary education institutions in the public sector; the rates within these organizations ranged from 21.0% for the men in one organization to 43.4% for the women in the other organization. By far, the highest response rates were achieved in the private sector organizations and the Crown corporation; the rates in these organizations ranged from 50.0% for the women in the equipment sales and service company to 100.0% for the men in that same company.  2. Description of the Sample In this subsection, the sample of respondents is described in terms of age, organizational tenure, education, job category, marital status, and ethnic designation. In addition, the sample of respondents (257 women, 197 men) is compared to the sample of potential respondents (725 women, 731 men) on the variables of age and organizational tenure. The sample of potential respondents includes all persons (in all seven participating organizations) to whom the questionnaire packages were distributed. Tables for all of these descriptive variables are located in Appendix B. Female and male respondents in both the actual and potential samples were very similar in age and organizational tenure (see Table B-2). The only difference noted was that the actual respondents were slightly younger than the potential respondents: the average ages in the actual sample were 35.62 years (for women) and 36.13 years (for men) as compared with 37.43 years (for women) and 37.46 years (for men) in the potential sample. The average lengths of organizational tenure in the actual sample were 6.00 years for women and 5.38 years for men; the means in the potential sample were 5.43 years for women and 5.86 years for men. The highest levels of education obtained by the actual respondents are outlined in Table B3. The percentages indicate that the women and men in the study had obtained roughly the same  128 levels of education. The majority of the respondents had obtained an undergraduate or graduate degree (67.7% of the women, 68.5% of the men). The next largest percentage of respondents had received a college diploma or had undertaken some college, university, or vocational training (26.0% of the women, 29.5% of the men). Finally, 6.2% of the women and 2.0% of the men were educated up to the level of high school graduation. The job categories within which the respondents were employed are listed in Table B-4. The single job category occupied by the largest percentage of respondents was that of professional (32.3% of the women, 30.5% of the men). This was followed by the technical job category which was occupied by 19.8% of the women and 21.8% of the men. The percentages of respondents who had indicated that they occupied some type of middle or senior management position were 20.3% for women and 21.9% for men. First line supervisor positions were occupied by 11.7% of the women and 6.7% of the men. In summary, the women and men respondents in this study were very similar in terms of the job categories in which they were employed. Table B-5 contains information about the marital status of respondents. Relatively equal percentages of women and men were contained within each category for marital status. The largest percentage of the sample was married (52.9% of the women, 60.9% of the men), followed by 25.7% of the women and 23.4% of the men in the never married or single category. The sample was predominantly Anglo-European/Canadian in terms of ethnic designation: 80.9% of the women and 85.8% of the men reported that they were Anglo-European/Canadian; much smaller percentages of respondents reported membership in other ethnic groups (see Table B-6).  B. HYPOTHESIS-TESTING ANALYSES The results of the statistical analyses performed to test the hypotheses associated with each of the five research questions addressed in the current study are presented in this section. The hypotheses are restated within each subsection as the findings associated with each research question are presented.  129  1. Research Question 1: What is the relationship between an individual's homosocial attitudes and the size of his or her same- and cross-gender networks?  Methods of Statistical Analysis  Two methods of statistical analysis were employed to evaluate the two hypotheses associated with Research Question 1. First, standard multiple regression was utilized to determine if size of same- and cross-gender supportive (mentoring) networks could be predicted from three homosocial attitudes (for men) and five homosocial attitudes (for women). The three homosocial attitudes common to women and men were Cross-Gender Preference, Same-Gender Preference, and Valuing. The two additional homosocial attitudes for women were Jealousy and Identification. (See Appendix C [Preliminary Analyses] for the results of the factor analyses of the homosocial attitudes data collected for the current study, and for a listing of the exact items used in computations of homosocial attitude subscale scores in the present study.) Separate regression analyses were performed for women and men (because of the different numbers of independent variables for each gender) and for each dependent variable. To counteract the increased risk of Type I error caused by the number of regression analyses (four), alpha was set at .0125 using the Bonferroni inequality procedure. As a second method of evaluating the hypotheses, one-tailed tests of significance were performed on the bivariate correlations between each independent variable and each dependent variable. Given the exploratory nature of the analyses regarding homosocial attitudes, the alpha for these tests of significance remained at .0125 so as not to increase unduly the probability of making Type II errors. As is described in more detail after the statement of each hypothesis, the results of the regression analyses did not produce particularly helpful information with which to make decisions about accepting or rejecting the hypotheses. Therefore, the results of the regression analyses are reported only in the text, and primarily for descriptive purposes: the exact values of nonsignificant  130 results are not reported; only those multipleftsor beta weights significant at the required alpha level of .0125 are reported. Acceptance or rejection of the hypotheses was based primarily on the results of the one-tailed tests of significance of the bivariate correlations between each independent variable (one of three to five homosocial attitudes) and each dependent variable (size of same-gender network or size of cross-gender network). These bivariate correlations are reported in Table 1 and described in the text which follows.  Hypotheses Hypothesis 1. The same-gender networks of respondents who report low levels of Jealousy (women only) and Cross-Gender Preference and high levels of Identification (women only), Same-Gender Preference, and Valuing will be significantly larger than the same-gender networks of respondents who report high levels of Jealousy (women only) and Cross-Gender Preference and low levels of Identification (women only), Same-Gender Preference, and Valuing.  The size of each respondent's same-gender network was determined by tallying the number of same-gender persons listed in the Network Table in which providers of supportive (mentoring) functions were listed. Standard multiple regression was performed between size of same-gender network (Size SGnet) as the dependent variable and the three (for men) and five (for women) homosocial attitudes as independent variables. The multipleftobtained for women was significant [ft = .26, F (5, 216) = 3.15, p < .01, n = 222]. However, none of the five homosocial attitudes had a significant beta weight. Two of the five bivariate correlations between Size SGnet and each of the five homosocial attitudes were significant in the hypothesized direction at the desired alpha level (.0125): the correlation between Cross-Gender Preference and Size SGnet was -.21 (p = .001); the correlation between Valuing and Size SGnet was .18 (p < .005). The remaining three bivariate correlations for women were not significant at the required alpha level. For the men, significance was not attained at the desired alpha level for the multipleft,the three beta weights, or the three bivariate correlations.  131 Table 1 Bivariate Correlations Between Homosocial Attitudes and Size of Same- and Cross-Gender Supportive (Mentoring) Networks  Predictor variable' Outcome variable  CG Pref  SG Pref  Valuing  Identification Jealousy  Women (n = 222) SizeSGnet  -.21***  SizeCGnet  .18**  .15* -.08  .18** -.08  .12* -.02  -.08 .00  Men (n = 168)  SizeSGnet  -.14*  SizeCGnet  .03  .02 -.10  .14* .13  Note. CG Pref = Cross-Gender Preference; SG Pref = Same-Gender Preference; Size SGnet = size of same-gender supportive (mentoring) network; Size CGnet = size of cross-gender supportive (mentoring) network. Probability levels of tests of significance are indicated down to the .05 level. However, because the alpha for tests of significance was set at .0125 using the Bonferroni inequality procedure, only those results marked with two or more asterisks reached the level of significance required for the current study. There were three predictor variables (homosocial attitudes) for men and five for women. The three homosocial attitudes common to men and women were Cross-Gender Preference, SameGender Preference, and Valuing. The two additional homosocial attitudes for women were Identification and Jealousy (because these two attitudes are not applicable to men, no bivariate correlations are reported for these variables for men). a  *p < .05.  **p < .005.  ***p< .001.  132 In summary, Hypothesis 1 was partially supported by the results for women and not supported by the results for men. For women, there was a significant relationship between the five homosocial attitudes in combination and the size of same-gender networks. The five nonsignificant beta weights after this significant multiple R likely were due to the shared variance of the independent variables. The clearest source of support for a portion of Hypothesis 1 was found in the women's bivariate correlations for Cross-Gender Preference/Size SGnet and Valuing/Size SGnet. As hypothesized, these results indicated that the same-gender networks of women respondents who reported low levels of Cross-Gender Preference and high levels of Valuing were significantly larger than the same-gender networks of women respondents who reported high levels of Cross-Gender Preference and low levels of Valuing. The women's results did not support the portions of Hypothesis 1 related to the homosocial attitudes of Same-Gender Preference, Identification, and Jealousy. For men, no relationship was found between homosocial attitudes and size of samegender networks.  Hypothesis 2. The cross-gender networks of respondents who report high levels of Jealousy (women only) and Cross-Gender Preference and low levels of Identification (women only), Same-Gender Preference, and Valuing will be significantly larger than the cross-gender networks of respondents who report low levels of Jealousy (women only) and Cross-Gender Preference and high levels of Identification (women only), Same-Gender Preference, and Valuing.  The size of each respondent's cross-gender network was calculated by counting the number of cross-gender persons listed in the Network Table as providers of supportive (mentoring) functions. Standard multiple regression was performed with size of cross-gender network (Size CGnet) as the dependent variable and the five (for women) and three (for men) homosocial attitudes as the independent variables. For women, there was not much relationship between the five homosocial attitudes in combination and size of cross-gender network: significance at the desired alpha level  133 (.0125) was not reached for the multiple R, and only Cross-Gender Preference had a significant beta weight [beta = .26, t (216) = 2.92, p < .005]. Only one of the five bivariate correlations was significant at the required level: the correlation between Cross-Gender Preference and Size CGnet was .18 (p < .005). For men, no relationship was indicated between homosocial attitudes and size of crossgender network. Significance was not reached for the multiple R, the beta weights, or the bivariate correlations. In summary, Hypothesis 2 was partially supported by the results for women and not supported by the results for men. The cross-gender networks of women who reported high levels of Cross-Gender Preference were significantly larger than the cross-gender networks of women respondents who reported low levels of Cross-Gender Preference. The homosocial attitudes of Same-Gender Preference and Valuing were not significantly related to size of cross-gender networks for women or men, nor were the attitudes of Identification and Jealousy (for women) and CrossGender Preference (for men).  2. Research Question 2: What is the relationship between an individual's homosocial attitudes and the levels of supportive (mentoring) behaviours and relationship provisions he or she receives from same- and cross-gender networks?  Methods of Statistical Analysis The two hypotheses associated with Research Question 2 were evaluated using the same methods of statistical analysis as those employed for Research Question 1. First, standard multiple regression was performed to assess whether levels of supportive (mentoring) behaviours and relationship provisions received from same- and cross-gender networks could be predicted from five homosocial attitudes (for women) and three homosocial attitudes (for men). Again, the three homosocial attitudes common to women and men were Cross-Gender Preference, Same-Gender Preference, and Valuing. The two homosocial attitudes applicable to women only were Jealousy  134 and Identification. (See Appendix C [Preliminary Analyses] for the results of the factor analyses of the homosocial attitudes data collected for the present study, and for a listing of the exact items used in computations of homosocial attitude subscale scores in the current study.) The levels of supportive (mentoring) behaviours received from same- and cross-gender networks were assessed by the Total Supportive (Mentoring) Behaviour Score (TSBS) for a) each respondent's same-gender network and b) each respondents cross-gender network. Relationship provisions were measured by scores (for same- and cross-gender networks) on four subscales of the Close Social Relationships Scale: a) Intimacy, b) Similarity, c) Defiance of Convention, and d) Respect for Differences. (A listing of the exact items used to compute these four subscale scores is contained in Appendix C, where the results of the factor analyses of the relationship provisions data collected for the current study are presented.) Separate regression analyses were performed for each of the five dependent variables. As well, separate analyses were conducted for women and men and for the provisions of same- and cross-gender networks. Therefore, because a total of 20 regression analyses were performed for this question, alpha was set at .0025 using the Bonferroni inequality procedure. The second method of evaluating the hypotheses associated with Research Question 2 was performing one-tailed tests of significance on the bivariate correlations between each predictor variable and each dependent variable. The alpha for these significance tests remained at .0025 so as not to increase unduly the risk of making Type II errors in this exploratory investigation. As with the results for Hypotheses 1 and 2 associated with Research Question 1, the results of the regression analyses did not produce particularly useful information with which to make decisions about accepting or rejecting Hypotheses 3 and 4. Again, the results of the regression analyses are reported only in the text, and primarily for descriptive purposes; exact values are reported only for those multiple Rs and beta weights that reached significance at the required alpha level of .0025. Acceptance or rejection of Hypotheses 3 and 4 was based primarily on the results of the one-tailed tests of significance of the bivariate correlations between each independent variable (one of three to five homosocial attitudes) and each dependent variable (levels of supportive  135 behaviours or relationship provisions received from same- and cross-gender networks). The bivariate correlations for women are reported in Table 2; the bivariate correlations for men are reported in Table 3.  Hypotheses Hypothesis 3. Respondents who report low levels of Jealousy (women only) and CrossGender Preference and high levels of Identification (women only), Same-Gender Preference, and Valuing will receive significantly higher levels of supportive (mentoring) behaviours and relationship provisions from same-gender networks than will respondents who report high levels of Jealousy (women only) and Cross-Gender Preference and low levels of Identification (women only), SameGender Preference, and Valuing.  Results for supportive (mentoring) behaviours. Separate standard multiple regression analyses were performed for women and men. For women, the TSBS for same-gender (SG) networks was regressed onto the five homosocial attitudes; for men, the SG TSBS was regressed onto the three homosocial attitudes. In addition, one-tailed tests of significance were performed on the bivariate correlations between each predictor variable and each dependent variable (see Tables 2 and 3). For women, significance at the required alpha level (.0025) was not attained for the multiple R, the five beta weights, or the five bivariate correlations. For men, the multiple R was not significant at the required alpha level (.0025). None of the three beta weights or the three bivariate correlations was significant at the required level either. In summary, the supportive (mentoring) behaviours portion of Hypothesis 3 was not supported by the results for women or men; no significant relationship was found between homosocial attitudes and the level of supportive (mentoring) behaviours received from same-gender networks.  136 Table 2 Bivariate Correlations Between Homosocial Attitudes and Levels of Supportive (Mentoring) Behaviours and Relationship Provisions Received From Same- and Cross-Gender Networks (Women)  Predictor variable Outcome variable  n  CG Pref  SG Pref  Valuing  Identification Jealousy  Same-gender networks .10  .14*  .09  25*****  .09  24*****  2 7 * * * * *  176  .03  .09  .15*  .14*  .08  DefCon  176  -.04  24*****  .20***  .10  .18^  RespDiff  175  2 g * * * * »  .20***  TSBS  188  Intimacy  176  Similarity  -.11  35*****  -.05  .01 - .11  -.22  Cross-gender networks TSBS  196  2Q****  -.14*  -.17**  Intimacy  153  .03  -.12  Similarity  153  .13  DefCon  153  .02  RespDiff  153  - .14*  -.07  .00  .01  .04  .03  -.05  - .06  .05  .08  .04  - .04  - .08  .13*  .10  .03  -.19**  21***  Note. CG Pref = Cross-Gender Preference; SG Pref = Same-Gender Preference; TSBS = Total Supportive (Mentoring) Behaviour Score; DefCon = Defiance of Convention; RespDiff = Respect for Differences. Probability levels of tests of significance are indicated down to the .05 level. However, because the alpha for tests of significance was set at .0025 using the Bonferroni inequality procedure, only those results marked with four or more asterisks reached the level of significance required for the current study. *p < .05. **p< .01. ***p< .005. ****p < .0025. *****p< .001.  Table 3 Bivariate Correlations Between Homosocial Attitudes and Levels of Supportive (Mentoring) Behaviours and Relationship Provisions Received From Same- and Cross-Gender Networks (Men)  Predictor variable Outcome variable  n  CG Pref  SG Pref  Valuing  Same-gender networks It  TSBS  159  -.14  .15*  .20**  Intimacy  142  -.07  .02  .09  Similarity  142  -.03  22***  .21**  DefCon  142  -.05  .12  .19*  RespDiff  142  - .10  .07  28****  Cross-gender networks 113  .04  Intimacy  87  -.19*  -.10  Similarity  87  .05  .11  DefCon  87  .11  .06  .23*  RespDiff  87  - .27**  3 4 * * * *  .17  TSBS  .07  .12  * .21  Note. CG Pref = Cross-Gender Preference; SG Pref = Same-Gender Preference; TSBS = Total Supportive (Mentoring) Behaviour Score; DefCon = Defiance of Convention; RespDiff = Respect for Differences. Probability levels of tests of significance are indicated down to the .05 level. However, because the alpha for tests of significance was set at .0025 using the Bonferroni inequality procedure, only those results marked with four asterisks reached the level of significance required for the current study. *p < .05.  **p < .01.  ***p < .005.  ****p< .001.  138 Results for relationship provisions. Separate standard multiple regression analyses were performed for women and men. For women, levels of Intimacy (INT), Similarity (SIM), Defiance of Convention (DC), and Respect for Differences (RD) received from same-gender networks were regressed separately onto the five homosocial attitudes. For men, levels of INT, SIM, DC, and RD received from same-gender networks were regressed separately onto the three homosocial attitudes. One-tailed tests of significance also were performed on the bivariate correlations between each predictor variable and each dependent variable (see Tables 2 and 3). For women, two of the multiple Rs were significant at the required alpha level (.0025): with Intimacy as the dependent variable, R = .38, F(5,170) = 5.67, p < .001; with Respect for Differences as the dependent variable, R = .40, F (5,169) = 6.32, p < .001. The multiple R for Defiance of Convention as the dependent variable just missed significance [R = .32, F(5,170) = 3.79, p < .005]. Despite these two significant multiple Rs and one nearly significant multiple R, none of the five beta weights following each multiple R was significant at the required alpha level. For women, many of the bivariate correlations between the predictor variables and the dependent variables were significant at the required alpha level. For Intimacy received from samegender (SG) networks as the dependent variable, three bivariate correlations were significant in the hypothesized direction at the required alpha level: for Cross-Gender Preference (CGP)/SG INT, r = -.25, p < .001; for Valuing (VAL)/SG INT, r = .34, p < .001; for Identification (ID)/SG INT, r = .27, p < .001. When Similarity with same-gender networks was the dependent variable, none of the bivariate correlations was significant at the required alpha. For Defiance of Convention with same-gender networks as the dependent variable, one bivariate correlation was significant in the hypothesized direction at the required alpha level (for SGP/SG DC, r = .24, p = .001). Finally, when Respect for Differences from same-gender networks was the dependent variable, three of the bivariate correlations were significant in the hypothesized direction at the required alpha level (for CGP/SG RD, r = -.35, p < .001; for VAL/SG RD, r = .26, p < .001; for JEAL/SG RD, r = -.22, p < .0025).  139 None of the four multiple Rs computed for men and none of the three beta weights following each multiple R was significant at the required alpha level of .0025. Only one of the bivariate correlations between the predictor variables and the dependent variables was significant in the hypothesized direction at the required alpha level: with Respect for Differences from same-gender networks (SG RD) as the dependent variable, the bivariate correlation between Cross-Gender Preference and SG RD was -.28 (p < .001). In summary, these results partially supported the relationship provisions portion of Hypothesis 3. For women's same-gender relationships, significant relationships were found between the five homosocial attitudes in combination and the relationship provisions of Intimacy and Respect for Differences. None of the five homosocial attitudes had significant beta weights following these significant multiple Rs, most probably due to the shared variance of the five homosocial attitudes. When the bivariate correlations between each homosocial attitude and each of the four relationship provisions were tested for significance separately, many significant relationships were found and all were significant in the hypothesized direction. The homosocial attitudes that were the best predictors varied according to the relationship provision. Women respondents who reported low levels of Cross-Gender Preference and high levels of Valuing and Identification received significantly higher levels of Intimacy from same-gender networks than women respondents who reported high levels of Cross-Gender Preference and low levels of Valuing and Identification. Women respondents who reported low levels of Jealousy and Cross-Gender Preference and high levels of Valuing received significantly higher levels of Respect for Differences from same-gender networks than women respondents who reported high levels of Jealousy and Cross-Gender Preference and low levels of Valuing. Women respondents who reported high levels of Same-Gender Preference received significantly higher levels of Defiance of Convention from same-gender networks than women respondents who reported low levels of Same-Gender Preference. No significant bivariate relationships were found (at the required alpha level of .0025) between homosocial attitudes and the relationship provision of Similarity.  140 For men's same-gender relationships, no significant relationships were found at the required alpha level between the three homosocial attitudes in combination and each of the four relationship provisions, and only one of the bivariate correlations was significant. For men, then, only one part of the relationship provisions portion of Hypothesis 3 was supported: men respondents who reported low levels of Cross-Gender Preference received significantly higher levels of Respect for Differences from same-gender networks than did men respondents who reported high levels of Cross-Gender Preference.  Hypothesis 4. Respondents who report high levels of Jealousy (women only) and CrossGender Preference and low levels of Identification (women only), Same-Gender Preference, and Valuing will receive significantly higher levels of supportive (mentoring) behaviours and relationship provisions from cross-gender networks than will respondents who report low levels of Jealousy (women only) and Cross-Gender Preference and high levels of Identification (women only), SameGender Preference, and Valuing.  Results for supportive (mentoring) behaviours. Standard multiple regression analyses were performed separately for women and men. For women, the TSBS for cross-gender (CG) networks was regressed onto the five homosocial attitudes; for men, the TSBS for cross-gender (CG) networks was regressed onto the three homosocial attitudes. As well, one-tailed tests of significance were performed on the bivariate correlations between each of the homosocial attitudes (predictor variables) and each dependent variable (see Tables 2 and 3). For women, significance at the desired alpha level (.0025) was not attained for the multiple R or for any of the five beta weights. Only one of the five bivariate correlations between each of the five homosocial attitudes and the TSBS for cross-gender networks was significant at the required alpha level (for CGP/CG TSBS, r = .20, p < .0025).  141 For men, significance was not attained at the desired alpha level (.0025) for the multiple R, the three beta weights, or the three bivariate correlations between each of the three homosocial attitudes and the TSBS for cross-gender networks. In summary, the supportive (mentoring) behaviours portion of Hypothesis 4 was partially supported by the results for women and not supported by the results for men. The one significant (at the .0025 alpha level) bivariate correlation (for women's CGP/CG TSBS) provided support for one portion of Hypothesis 4: women respondents who reported high levels of Cross-Gender Preference received significantly higher levels of supportive (mentoring) behaviours from cross-gender networks than women respondents who reported low levels of Cross-Gender Preference. For women, significant relationships were not found between the homosocial attitudes of Same-Gender Preference, Valuing, Identification, and Jealousy and the levels of supportive (mentoring) behaviours received from cross-gender networks. The results for men indicated that there was no significant relationship between homosocial attitudes and the level of supportive (mentoring) behaviours received from cross-gender networks.  Results for relationship provisions. Standard multiple regression analyses were performed separately for women and men. For women, levels of Intimacy (INT), Similarity (SIM), Defiance of Convention (DC), and Respect for Differences (RD) received from cross-gender (CG) networks were regressed separately onto the five homosocial attitudes. For men, levels of INT, SIM, DC, and RD received from cross-gender networks were regressed separately onto the three homosocial attitudes. In addition, one-tailed tests of significance were performed on the bivariate correlations between each homosocial attitude and each of the four relationship provisions (see Tables 2 and 3). For women, none of the four multiple Rs, the beta weights, or the bivariate correlations was significant at the required alpha level of .0025. For men, two of the multipleflswere significant at the required alpha level of .0025 [for Intimacy as the dependent variable, R = .45, F(3,83) = 7.08, p < .001; for Respect for Differences as the dependent variable, R = .43, F (3,83) = 6.41, p < .001]. Only one of the beta weights following  142 the significant multiple R for Intimacy was significant. However, ft was significant in the opposite direction to that hypothesized [the beta weight for VAL was .46, t (83) = 4.16, p < .001 ]. The bivariate correlation between VAL and CG INT also was significant in the opposite direction to that hypothesized (r = .40, p < .001). Neither of the remaining two bivariate correlations for Intimacy was significant at the required alpha level. One of the five beta weights following the significant multiple R for Respect for Differences was significant in the hypothesized direction at the required alpha level [the beta weight for SGP was -.37, t (83) = -3.34, p < .0025]. The bivariate correlation between SGP and CG RD also was significant in the hypothesized direction (for SGP/CG RD, r = -.34, p = .001). Neither of the remaining two bivariate correl