Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Saying yes, saying no : understanding women’s use of the label "PMS" Moore, Shelley 1997

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

Item Metadata

Download

Media
831-ubc_1998-272044.pdf [ 17.77MB ]
Metadata
JSON: 831-1.0088767.json
JSON-LD: 831-1.0088767-ld.json
RDF/XML (Pretty): 831-1.0088767-rdf.xml
RDF/JSON: 831-1.0088767-rdf.json
Turtle: 831-1.0088767-turtle.txt
N-Triples: 831-1.0088767-rdf-ntriples.txt
Original Record: 831-1.0088767-source.json
Full Text
831-1.0088767-fulltext.txt
Citation
831-1.0088767.ris

Full Text

SAYING Y E S , SAYING NO: UNDERSTANDING WOMEN'S USE OF THE LABEL " P M S " by SHELLEY MOORE B.A., The University of Windsor, 1984 M.A., Laurentian University, 1990 A DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE D E G R E E OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Psychology) We accept this dissertation as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA November 1997 ©Shelley Moore 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 The University of British Columbia Vancouver, Canada DE-6 (2/88) I Abstract This dissertation explored factors related to heterosexual women's use of the label "PMS" and the question of why so many women who say that they have PMS do not meet prospectively-based researcher criteria. Participants were recruited through Vancouver city newspapers and posters for a study of emotional, behavioural, and health patterns. They were screened for hysterectomies, ovariectomies, pregnancy, and chronic illnesses. The 58 women in this study (mean age=34.5) provided daily prospective data over a mean of 15 weeks for 15 variables representing 4 different types of symptoms: mood, relationship, concentration and work performance, and physical and activity symptom types. At the beginning of the study, they completed the Beck Depression Inventory. During a final interview they completed various questionnaires regarding romantic relationships, stress, history of abuse, and attitudes toward menstruation. During her final interview, each woman was asked whether she had ever had PMS and, if so, what she believed caused it. Only 16% of the women met researcher-designated criteria for PMS and 9% met researcher-criteria for PDD (Premenstrual Dysphoric Disorder, APA, 1994), but 60% believed that they currently had PMS. There was very little match between researcher-designations and self-designations for any of the 15 prospective variables. Instead, consistent with schematic theories of PMS, it was women's attitudes toward menstruation that were most strongly related to self-designation. Women who said that they currently had PMS were more likely to view menstruation as debilitating, natural, and predictable. They had higher depression scores and reported more frequent and enduring experiences of anger. More of these women reported having been emotionally abused as an adult, emotionally abused as a child, and physically abused by a past or current romantic partner. Although women who said that they had never had PMS Ill perceived menstruation as more bothersome, they believed that women should be able to ignore it. Current-Say and Never-Say women did not significantly differ for other forms of abuse, partner satisfaction, or daily uplifts. Daily hassles did not reach statistical significance at the multivariate level, but univariate significance indicated that Never-Say women might experience fewer daily hassles than Current-Say women. The prospective data were analyzed idiographically using prediction analyses. Current-Say women demonstrated more uncharacteristic cyclicity during the midcycle phase across the 15 variables and 4 symptom types. Never-Say women showed more uncharacteristic cyclicity during the postmenstrual phase. No differences were found for other phases. These findings, and other results, argue against the use of simple premenstrual-postmenstrual and premenstrual-intermenstrual difference measures in diagnosing PMS or PDD. It was concluded that, although the women's self-designations did not match their prospective data, they could not be explained simply as a mislabelling of negative cyclicity in other phases. There was also mixed evidence for the hypothesis that women's reports of having "PMS" were part of a generalized over-reporting of symptomatology or dissatisfaction. Possible explanations for women's self-designations are discussed, including schematic representations of both menstruation and general illness and a "contrast hypothesis" modified from the version proposed by McFarlane and Williams (1994). This contrast hypothesis suggests that cyclical changes occurring during a particular non-premenstrual phase are related to women's self-designations according to (a) the timing of these changes relative to the visible and salient menses, in conjunction with (b) their attitudes toward menstruation. Close to half (45%) the women who said that they had experienced PMS attributed their perceived PMS to a bidirectional model of physiology and environment, and 58% of the women saying that they had PMS believed that it happened occasionally rather than inevitably. Overall, the women's representations of menstrual cyclicity were neither simple reflections of cultural stereotypes nor pervasively negative, but rather diverse and complex. The results that link depression, anger, and specific forms of abuse to self-designations suggest that women saying that they have PMS are indicating difficulties that may or may not be menstrually-related. Researchers and other professionals need to assess the nature of those difficulties in women presenting with "PMS". V Table of Contents Abstract ii List of Tables xii List of Figures xvi Acknowledgements xviii Introduction 1 Defining PMS 1 The Influence of Stereotypes on Women's Reports of PMS 3 Are Women Reporting "PMS" Retrospectively? 6 Retrospective Symptom Reports of Women Who Say They Have PMS ...6 Treatment-seeking for PMS 8 "PMS" as an Illness Label 9 The Role of Stress in Understanding PMS 10 The Role of Relationships in PMS 14 Abuse in Relationships and PMS 17 PMS and Depression 18 Variability Among Women 19 Purposes of this Study 22 Method 24 Participants 24 Procedure 27 Initial interview 27 vi Daily questionnaire 27 Final interview 29 Researchers 31 Self-designation of "PMS" 31 Coding of Women's Causal Explanations for PMS 33 Procedure 33 Coding system 34 Definitions of Menstrual Phases 34 Analysis of Daily Variables 35 Idiographic analyses 35 The criteria for research designations of PMS and PDD 36 Analysis of Single-measurement Variables 37 Some methodological errors to consider 37 Results 40 Approach to Normative Data Analysis 40 Type I and Type II Error 47 Say Group Analyses of Single-measurement Data 52 Do Women's Reports About their Life Circumstances, Emotions, and Attitudes Relate to Their Current Use of the Label "PMS"? 52 Summary of the single-measurement findings 67 Say Group Normative Analyses of Prospective Data 68 Analyzing Symptom Types 68 vii Are Women's Current Self-Designations of "PMS" Related to Changes They Experience on Specific Types of Symptoms? 69 Examining How Women Who Say They Have PMS Are Using the Label 77 Do women's reports about their life circumstances, emotions, and attitudes relate to the consistency with which they use "PMS"? 77 Does the consistency with which women use "PMS" relate to their prospective cyclicity? 89 Say Group Idiographic Analyses of Prospective Data 89 Were Women's Self-designations of PMS Related to Researcher-designations of PMS? 89 A comparison of PMS researcher-designations and PMS self-designations 92 PDD and nonPDD classifications 98 PMS, PDD, and Say group classifications for women taking oral contraception and for women with clinical BDI scores 100 Were women's self-designations related to the amount of cyclicity they experienced? 100 Were women's self-designations related to the distribution of cyclical clusters over their menstrual cycles? 104 Were women's self-designations related to the number of daily variables they experienced over the menstrual cycle? 113 Were women's self-designations related to a more sensitive detection of cyclicity? 116 Summary of T, M, and Cyclicity Findings 121 Analyses of Current-Say and Never-Say nonPMS Women 125 Causal Attributions of Women Who Said That They Have Experienced PMS 130 viii How Did Women Who Said They Had PMS Explain Their Own Perceived PMS? 130 Were Women's Explanations of Their Own PMS Related to Their Attitudes, Emotions, and Life Circumstances? 133 Were Women's Explanations of Their Own PMS Related to Their Prospective Patterns for Symptom Types? 141 Were the Causal Explanations Provided by Women Who Said That They Had PMS Related to Researcher Designations?..141 Were Women's Explanations of Their Own PMS Related to the Amount of Cyclicity They Experienced Over the Menstrual Cycle Phases? 144 Summary of Findings Associated With Women's Causal Explanations of PMS 147 A Review of the Findings 148 Discussion 163 Did Women's Self-designations Match Their Prospective Cyclicity? 163 Do Women's Self-designations Reflect Cultural Stereotypes About PMS? 165 Are Women's Self-designations Related to Stress? 166 Contributions From Illness Cognition Research 168 Is There Support for a Role Violation Hypothesis? 174 Is Self-designation as Having PMS Part of a Generalized Pattern of Over-reporting? 177 Are Women Mislabelling Patterns of Menstrual Cyclicity? 181 Evidence for PMS and PDD 185 Summary and Conclusions 190 ix Strengths, Limitations, and Directions for Future Research 194 References 200 Appendix A: Literature Review 233 Incidence of Premenstrual Syndrome 234 Defining PMS 235 What is Premenstrual? 235 Duration of the premenstrual phase 235 The timing of the premenstrual phase 236 Establishing the baseline for premenstrual comparison 238 Who Put the S in PMS? 241 Symptomatology of PMS 241 PMS as a syndrome 243 Naming premenstrual symptomatology 244 Standardizing the Definition of PMS 247 The National Institute of Mental Health criteria 247 Diagnostic and Statistical Manual-lll-R criteria 249 Diagnostic and Statistical Manual-IV criteria 252 Opposition to LLPDD and PDD 257 Method, Measures, and Assumptions in Determining PMS 261 Instruments to Assess PMS 261 The Menstrual Distress Questionnaire 262 The Premenstrual Assessment Form and Daily Rating Form ....264 X Other measurement instruments 265 Assuming a Biological Origin 267 Evidence for a biological model 268 Menstrual cyclicity as independent or dependent variable 273 Assuming the Nature of PMS 276 Assumption of adverse change 276 Assumptions about cyclicity 280 Publishing defines PMS 284 The Role of Stereotypes in PMS 285 Retrospective and Prospective Methods 285 The Role of Stereotypes in Understanding PMS 292 Treatment-seeking for PMS 299 The Discrepancy Between Women's and Researchers' Understanding of PMS ...301 The Role of Stress in Understanding PMS 305 The Role of Relationships in PMS 310 PMS and Depression 315 Multidimensionality and Variability in PMS 318 Appendix B: A Partial Chronological Tabular Review of PMS Articles 323 Appendix C: DSM-III-R Criteria for Late Luteal Dysphoric Disorder and DSM-IV Criteria for Premenstrual Dysphoric Disorder 342 Appendix D: Initial Interview Package 345 Appendix E: Daily Chart and Instructions for Completion 353 xi Appendix F: Final Interview Package 359 Appendix G: Reliability Study for New Measures and Reported Reliability of Published Measures 405 Appendix H: Diagnostic Manual for Premenstrual Syndrome Researcher Designations 419 xii List of Tables Table 1: Demographics of sample 25 Table 2: Correlations among single-measurement variables 44 Table 3: Results of parametric analyses of single-measurement variables for differences between Current-Say and Never-Say women 53 Table 4: Results of abuse variables for Current-Say and Never-Say women ...66 Table 5: Means and confidence intervals for the prospective variables for the Current-Say and Never-Say women 70 Table 6: Results of parametric analyses of prospective variables for the Current-Say and Never-Say women 71 Table 7: Means and confidence intervals for the single-measurement variables for Never-Say, Sometimes-Say, and Always-Say women ..78 Table 8: Results of parametric analyses of single-measurement variables for differences among Never-Say, Sometimes-Say, and Always-Say women 80 Table 9: Results of non-parametric analyses of abuse variables for Never-Say, Sometimes-Say, and Always-Say women 88 Table 10: Means and confidence intervals for the prospective variables for the Never-Say, Sometimes-Say, and Always-Say women 90 Table 11: Results of parametric analyses of prospective variables for the Never-Say, Sometimes-Say, and Always-Say women 91 Table 12: Women's self-designations of PMS, researcher-designations of PMS, and the distribution of terrific and miserable 3-item clusters by menstrual phase 93 Table 13: Women's self-designations of PMS, researcher-designations of PDD, and the distribution of terrific and miserable 5-item clusters by menstrual phase 95 Table 14: Number of women meeting researcher-designated criteria of PMS for each say group 97 xiii Table 15: Number of women meeting researcher-designated criteria of PDD for each say group 99 Table 16: Distribution of terrific and miserable clusters for Current-Say and Never-Say self-designations 101 Table 17: Distribution of terrific and miserable clusters for Always-Say, Sometimes-Say, and Never-Say self-designations 102 Table 18: Distribution of diagnosable M clusters for Current-Say and Never-Say self-designations 105 Table 19: Distribution of diagnosable M clusters for Always-Say, Sometimes-Say and Never-Say self-designations 108 Table 20: Distribution of diagnosable and nondiagnosable M clusters for Current-Say and Never-Say self-designations 110 Table 21: Distribution of diagnosable and nondiagnosable M clusters for Always-Say, Sometimes-Say, and Never-Say self-designations 111 Table 22: Distribution of diagnosable T clusters for Current-Say and Never-Say self-designations 112 Table 23: Distribution of diagnosable T clusters for Always-Say, Sometimes-Say, and Never-Say self-designations 114 Table 24: Mean number of uncharacteristically cyclical variables in each menstrual phase for Current-Say and Never-Say self-designations 115 Table 25: Mean number of uncharacteristically cyclical variables in each menstrual phase for Always-Say, Sometimes-Say, and Never-Say self-designations 117 Table 26: Distribution of cyclicity based on a 1-item liberal criterion for Current-Say and Never-Say self-designations 118 Table 27: Distribution of cyclicity based on a 1-item liberal criterion for Always-Say, Sometimes-Say, and Never-Say self-designations 120 xiv Table 28: A summary of significant findings for statistical comparisons conducted on the idiographic data for Current-Say and Never-Say women 122 Table 29: Means and confidence intervals of single-measurement variables for nonPMS Current-Say and Never-Say women 126 Table 30: Results of single-measurement variables for nonPMS Current-Say and Never-Say women 127 Table 31: Percentages of nonPMS Current-Say and Never-Say women reporting abuse 128 Table 32: Causal attributions provided by Always-Say and Sometimes-Say women 134 Table 33: Means and confidence intervals for single-measurement variables for the cause-iv, cause-dv, and cause-both groups 135 Table 34: Results of parametric analyses of single-measurement variables for differences among cause groups 137 Table 35: Percentages of cause-iv, cause-dv, and cause-both women reporting history of abuse 140 Table 36: Means and confidence intervals for prospective variables for the cause-iv, cause-dv, and cause-both women 142 Table 37: Results of parametric analyses of prospective variables for cause-iv, cause-dv, and cause-both women 143 Table 38: Statistically significant non-idiographic findings for which there was strong evidence 149 Table 39: Statistically significant non-idiographic findings for which there was relatively moderate or weak evidence 150 Table 40: Statistically non-significant findings for non-idiographic analyses. 151 Table 41: Test-retest reliability and internal consistency calculations for new scales: Women and men 409 Table 42: Test-retest reliability and internal consistency calculations for new scales: Women 410 X V Table 43: Test-retest reliability and central tendency measures for questions included in new scales: Women and men 411 Table 44: Test-retest reliability and central tendency measures for questions included in new scales: Women 413 xvi List of Figures Figure 1: Menstrual Attitude Questionnaire Subscales on Which Never-Say Women Scored Significantly Higher than Current-Say Women 55 Figure 2: Menstrual Attitude Questionnaire Subscales on Which Current-Say Women Scored Significantly Higher than Never-Say Women 57 Figure 3: Anger and Depression Scales on Which Current-Say Women Scored Significantly Higher than Never-Say Women 59 Figure 4: Anger and Sadness Scales for Which No Significant Differences were Found Between Current-Say and Never-Say Women 61 Figure 5: Responses Regarding Current Romantic Relationships for Which No Significant Differences were Found Between Current-Say and Never-Say Women 63 Figure 6: Daily Hassles and Uplifts Scores of Current-Say versus Never-Say Women 64 Figure 7: Results of Prospective Analyses Across Say Groups: Menstrual Phase Comparisons for Relationship Variables and for Concentration and Work Performance Variables 72 Figure 8: Results of Prospective Analyses Across Say Groups: Menstrual Phase Comparisons for Mood and for Physical and Activity Variables 73 Figure 9: Results of Prospective Analyses Across Say Groups: Symptom Type Comparisons for the Menstrual and Premenstrual Phases 74 Figure 10: Results of Prospective Analyses Across Say Groups: Symptom Type Comparisons for the Postmenstrual, Midcycle, and Post-midcycle Phases 75 Figure 11: Subscales of the Menstrual Attitude Questionnaire for Which 3 Say Group Findings were Different from 2 Say Group Findings 83 xvii Figure 12: Depression Scores for the Never-Say, Always-Say, and Sometimes-Say Women 85 Figure 13: Causal Attributions for Women Who Say That They Have PMS 132 Acknowledgements I would like to acknowledge and extend appreciation to the many individuals who have directly and indirectly contributed to the production of this dissertation: First, this dissertaton was completed on Musqueam land. I would like to recognize and thank the land and its guardians. Second, I offer my sincerest appreciation to my family for their commitment, faith, and patience. Most especially, this degree is shared by my parents, whose love and undying support have made the effort worthwhile. Third, I owe the completion of this degree to the many sacrifices of my loving life partner, Squeig. Her continued nurturance, understanding, enthusiasm, faith, and words of inspiration have sustained me throughout this process. Fourth, I would like to thank my supervisor, Tannis MacBeth, and my committee members, Anita DeLongis and Larry Walker, for their efforts, helpful comments, and encouragement. In addition, I wish to extend my appreciation to the many workers of UBC who have made my own work possible. In particular, I am grateful for the efforts of the many service workers, maintenance and cleaning workers, and clerical workers on campus, as well as the staff of Graduate Studies and the libraries. Max Bhatti has contributed extensive effort and expertise in producing impeccable duplications of my dissertation over the past two years. I have also received helpful advice and consultation from Janet Werker, Geoff Hall, Mark Schaller, Linda Scratchley, Kim Berchard, and Shawn Reynolds. There have been numerous lab assistants who have participated in this project since its beginning. I would like to thank these individuals and, in particular, share the success of the study with Tami Nicholson, Evelyn Dalian, and Karla Dye-Tiffoli. Vital to my pursuit of this topic and this project has been the constant encouragement of Jessica McFarlane, my colleague and predecessor. And, although their names do not appear here, I would like to recognize the extensive efforts of the many women who provided data for this study. xix I would also like to honour the many friends and companions who have offered their time, energy, and ear: Indy Batth, Anne Beaulne, Imogene and Roy Blunden, Darlene Brodeur, Cynthia Brookes and Mariana, Chris Bruels, Marianna Brussoni, Richard Buffalo, Benita Bunjun, Rodney DeCroo, Marg Dorey, Fireweed, GLOBAL, Leanne Hammond, Karl Henning, Jo Hinchliffe, Pat Hogan, Lyndsey Jennings, Angela Lamensdorf, Gitanjali Lena, LSU, Ann McCabe, Sue McGowan, Judy Mason, Kyle Matsuba, Ryan Mercredi and Sean Drygeese, Michelle Montgomery, Cleo Pawson, Phoenix, Kristin Schopp, Rima Sultan, Lome Sutherland, VWC, past and present WMST 300 students, WMST instructors (notably Gillian Creese, Dawn Currie, Yasmin Jiwani, Pat Kachuk, Ann MacKinnon, Marina Morrow, Becki Ross, Dorothy Seaton, Marni Stanley), and, of course, Terrapin, Darby, and Muncey. Most especially, I thank Diana, for Her continuing love, strength, compassion, and guidance. Finally, it is important to recognize that this dissertation cannot be decontextualized from the racialized and class-based structure of both the North American and global societies in which it was written. It is important to acknowledge that the privilege I am receiving from the bestowal of this degree is inseparable from the privilege that has been denied to others. Dedication This dissertation is dedicated to Dr. Michael Chapman, who I continue to deeply respect and miss. It is also submitted in memory of Pamela Leslie, Shirley (Mills) Whitesell, Lloyd Whitesell, Tyffin, Shadow, and an anonymous man whose accidental death touched my life. 1 Introduction Defining PMS One of the most pivotal yet unresolved issues in menstrual cycle research has been the definition of "Premenstrual Syndrome" (PMS). Reviewers have noted prevalence rates ranging from 5 to 95% (Ussher, 1992a), from 5 to 97% (Veeninga & Kraaimaat, 1985), from 20 to 90% (J. P. Siegel, 1986), and from 29 to 97% (Bancroft & Backstrom, 1985). Such broadly encompassing reports have resulted from widely differing ways of defining and measuring PMS. There has been little agreement among researchers on the number, timing, or duration of the phases of the menstrual cycle. The part of the cycle against which the premenstrual phase has been compared to assess PMS pathology has varied across studies. There have been no consistent criteria for distinguishing menstrual "symptoms" from normal menstrual "changes". Moreover, the literature lacks reliable standards for deciding on the type, severity, and number of symptoms required to constitute a "syndrome". There is not yet even agreement that PMS either is or should be a syndrome. In an effort to incorporate the disparate approaches to PMS, in the literature of the last decade PMS has increasingly been represented as a multidimensional construct (Woods, Lentz, Mitchell, & Kogan, 1994). The lack of agreement on a core set of symptoms has led some researchers to re-emphasize Moos' (1969) original proposal that PMS is a composition of "syndromes", rather than a unitary "syndrome" (Abplanalp, 1983; Bancroft & Backstrom, 1985; Cumming, Cumming, Krausher, & Fox, 1991; Halbreich, Endicott, & Nee, 1983; McMillan & Pihl, 1987; Reid, 1985; Rivera-Tovar, Pilkonis, & Frank, 1992; Ussher, 1991; Van der Ploeg & Lodder, 1993) (For a more thorough description of PMS definitions as documented in the research, see Appendix A). Moreover, these and other efforts have been pursued primarily to establish the meaning of PMS among 2 researchers. Relatively little research has addressed how the women who are the "objects" of this definitional discourse interpret "PMS". We collected data for a large longitudinal project to examine women's cyclical patterns.1 In the first study deriving from this project, no relationship was found (phi=-.08) between women's self-designations as having or not having PMS and these same women's prospectively determined PMS status based on 4 months of daily mood data (McFarlane & Williams, 1994). This suggests that women may be using the label "PMS" differently than researchers or may not be accurately recalling their menstrual patterns in applying the "PMS" label. Several reviewers have called for more direct investigation of how women who use the term PMS define it (Cumming, Urion, Cumming, & Fox, 1994; Koeske, 1980b; Rodin, 1992; Stoppard, 1992). To clarify communication about PMS between women and researchers, as 1 The choice of cross-sectional or longitudinal methodology is an issue of debate among developmental psychologists. McCall (1977) has referred to longitudinal research as the lifeblood of developmental psychology. Prospective (longitudinal) data are especially important to menstrual cycle research. In an effort to reduce cost, optimize time investment, and allow analysis of the full richness of longitudinal data, some reviewers have strongly recommended that a multipurpose approach be taken to permit the production of several smaller studies within the larger project (Harway, Mednick, & Mednick, 1984; Mednick, Griffith, & Mednick, 1981; Mednick, Mednick, & Griffith, 1981). Concerns regarding the accrual of Type I error often precludes or discourages the analysis of all potentially relevant variables and questions within a single study. Yet, without the analysis of additional information, a more complete picture as well as possible contradictory findings may be overlooked (Colby & Phelps, 1990). This longitudinal project is consistent with those recommendations. Two studies (McFarlane & Williams, 1994 and this dissertation) have been conducted on the same women, addressing different questions and analyzing different parts of the data set. Data collection was collaborative, but the studies were designed, analyzed, and produced independently. Some reviewers have strongly argued for an archival approach to maximize the value of longitudinal research, which requires that the researcher analyze the previously collected data of another researcher (Colby & Phelps, 1990). This dissertation was not an archival study, but rather part of a larger collaborative one, since the author participated fully from the early stages of the project (that is, prior to the application for funding and data collection). 3 well as between women and the professionals from whom they seek help for PMS, we must begin to explore how this self-labelling is meaningful to the individual women who use it (Gallant, Popiel, Hoffman, Chakraborty, & Hamilton, 1992b). This dissertation is one of the first to systematically examine factors related to women's choices to label themselves as having or not having "PMS". Early research focused on a biomedical model of PMS, but the results have been inconsistent (See Appendix A for a review of biological evidence). During the last decade, research has shifted increasingly toward a biopsychosocial approach. This dissertation was conducted within the psychosocial part of that realm. The Influence of Stereotypes on Women's Reports of PMS Previous efforts to study researcher and participant differences in defining PMS have focused on the well-documented discrepancy between women's retrospective and prospective reports of symptomatology (Alagna & Hamilton, 1986; Christensen & Oei, 1989; Clare, 1985; McFarlane, Martin, & Williams, 1988; McFarlane & Williams, 1994; Ussher, 1992b). Whereas 60 to 70% of women report premenstrual symptoms retrospectively, 20 to 50% demonstrate this pattern prospectively over one cycle and only 5% show prospective confirmation over two cycles (McFarlane & Williams, 1994). In contrast to the number of women who retrospectively report PMS, a cross-sectional epidemiological study conducted on a population-based random sample of 2650 urban Canadian women demonstrated a 1% prevalence rate of psychological symptoms occurring during the premenstruum (Ramcharan, Love, Fick, & Goldfien, 1992). In one study (Gallant, Popiel, Hoffman, Chakraborty, & Hamilton, 1992b), many women who rated their symptoms as severe did not even meet the most liberal criteria for confirmation. Retrospective data have generally been considered by social scientists to be less reliable than prospective data (Dua, 1995; Harrison, Mullen, & Green, 1992; 4 Janson, 1990; Nollan, Kennedy, & Kennedy, 1991). In a study of seven different domains, Henry et al. (1994) found that psychosocial variables, such as subjective psychological states and family processes, were most affected by the choice of a retrospective or prospective approach. Others have documented the discrepancy between retrospective and prospective data in health research. Whereas negative affect has been shown to be the best predictor of retrospective health reports, attributional style has been found to be the best predictor of prospective health reports (Dua, 1995). In a meta-analysis conducted by Harrison et al. (1992), retrospective and prospective studies differed in the dimensions that most influenced health beliefs and behaviour. Retrospective reports about the frequency and timing of events are generally inaccurate (Henry et al., 1994). Errors are especially found in dating phases within a continuous process (Janson, 1990), indicating that the use of retrospective measures to assess menstrual phase experiences may be particularly inappropriate. The inaccuracy of retrospective reports has been attributed, in part, to forgetting (Henry et al., 1994). Accurate recollection decreases as the time span being recalled increases (Janson, 1990). This effect has been attributed to the proactive and retroactive interference of other experiences (Janson, 1990). In addition, individuals often re-structure memories of past events to suit their current context and understanding (Henry et al., 1994). As memory fades, recall is interpreted through beliefs, expectations, and general scripts of past experience (Janson, 1990). Patterns and details are changed to form a logical and coherent picture, consistent with this interpretive lens (Janson, 1990). Some reviewers have claimed that retrospective questionnaires in PMS research tend to elicit data that match cultural stereotypes of menstrual experience 5 rather than data on actual menstrually-related changes (Gannon, 1981; McFarlane & Williams, 1990; Ussher, 1991; Ussher, 1992a). The search for a theoretical framework that incorporates the role of culture in understanding menstrual cyclicity has led to a shift toward cognitive models (Ussher, 1992a). The foremost theorist in this area has been Koeske (Koeske & Koeske, 1975; Koeske, 1980a; 1980b; 1983; 1986). Her "arousability labelling hypothesis" is a schematic one which proposes that women interpret their physiological experiences based on situational and cultural cues. Consistent with this theory, person perception tasks have shown that negative premenstrual behaviour is noticed and understood as evidence for popular notions of PMS, but positive premenstrual behaviour is not linked to the biology of the menstrual cycle (Bains & Slade, 1988; Koeske & Koeske, 1975; Koeske, 1981, 1986). Moreover, when premenstrual symptoms are ascribed to biology, the situational factors that are simultaneously operating are discounted. In addition, biological explanations are less likely to be used to account for negative moods occurring in non-premenstrual phases. Further evidence for the role of expectations in symptom reporting derives from research in which women who were intentionally misinformed that they tested as hormonally premenstrual later reported having had more symptoms than did women in the same menstrual phase who were told that they tested as intermenstrual (Klebanov & Jemmott, 1992; Ruble, 1977). Other investigators have also shown a relationship between menstrual attitudes and retrospective reporting (Brooks-Gunn, 1985; Klebanov & Jemmott, 1992; Woods, Dery & Most, 1982b). Research in our own lab (McFarlane et al., 1988; McFarlane & Williams, 1994) has shown that women and men were also influenced by stereotypes regarding days of the week. They recalled "Monday blues" and "Saturday highs" patterns that were not prospectively confirmed. These findings 6 add support to the interpretation that retrospective reports may be biased by expectations. Are Women Reporting "PMS" Retrospectively? Retrospective Symptom Reports of Women Who Sav Thev Have PMS Analyses of mood data support the notion that women's self-designations of PMS cannot be inferred directly from their retrospective symptom reports (McFarlane & Williams, 1994). Only 76% of the women who said that they had PMS indicated such a pattern retrospectively and 43% who said that they did not have PMS provided retrospective data matching a PMS pattern. Warner, Bancroft, Dixson, and Hampson (1991) found that 95% (146 women) of their participants believed that they suffered from PMS. Even though the remaining 5% (8 women) said that they did not have PMS, most of these women retrospectively indicated PMS symptomatology. Ussher (1992a) found that women who self-diagnosed as having PMS did not differ from women who said they did not have PMS in either their retrospective or prospective symptom reports. This difference between retrospective reporting and women's self-diagnoses indicates that women's use of "PMS" is not simply reducible to menstrual stereotypes. The study of the retrospective-prospective discrepancy has been approached both as a measurement issue and as a barometer of menstrual stereotypes. The studies just reviewed suggest, however, that neither measure may be a complete or accurate reflection of how women are assessing PMS. A new question posed by this dissertation is, "What are women saying when they label themselves as having PMS?". This question is especially pertinent to Koeske's arousability labelling hypothesis. Without direct information on women's understandings of the label "PMS", it is difficult to project exactly what schemata of menstruation women may be using to interpret their experiences. Veeninga and Kraaimaat (1995) have noted that, to date, we know little about women's explanatory models for their own cyclical symptoms. Koeske (1980) has critiqued this lack of information about women's implicit self-theorizing. She has emphasized the failure of research to include women's personal and social interpretations and assumptions about the menstrual cycle as equally substantive theoretical variables. Such a neglect puts the emphasis on what the professional deems "problematic" but may exclude what the woman finds most troublesome. Women's definitions of what constitutes premenstrual syndrome are particularly relevant to the recent inclusion of Premenstrual Dysphoric Disorder (PDD) in the appendix of the DSM-IV (APA, 1994). Criterion B of this disorder requires that premenstrual disturbance seriously interferes with work, social activities, or relationships with others. Gallant and Hamilton (1988) have noted that this criterion is not operationalized, but is defined subjectively by the woman. Information is required about the extent of and types of premenstrual interference that women consider sufficiently serious to label a disorder or syndrome. 2 If the clinician is unaware of the patient's explanations of PMS, she or he may prescribe a treatment strategy that is not consistent with the woman's view of her own symptoms (Veeninga & Kraaimaat, 1995). Indeed, women's perceptions of their need for PMS treatment and professionals' perceptions of when women require treatment for PMS seem to differ. 2 For a thorough analysis of the PDD classification and controversy, see Appendix A. In particular, I have challenged the claim by the DSM authors that PDD and PMS are distinguishable based on the affective nature of the former. Only 1 symptom of 5 required from a list of 10 options must be affective for diagnosis. It is possible to be diagnosed with PDD based on 4 physical symptoms and only 1 emotional symptom. Stoppard (1992) has noted 186 possible symptom combinations which may lead to PDD diagnosis. 8 Treatment-seeking For PMS Treatment-seeking has been shown to be a poor predictor of diagnosed premenstrual distress. The discrepancy rate between treatment-seeking and prospective confirmation of PMS ranges from 10 to 40% (Ekholm & Backstrom, 1994). In one study, only a third of the women who presented at a clinic for PMS showed prospective evidence (Stout & Steege, 1985). Christensen and Oei (1995) found no difference between treatment-seekers and non-seekers in demographics or in number of symptoms, but the two groups varied in the nature of symptoms. Treatment-seekers were more symptomatic on depression, anxiety, frequency of negative automatic thoughts, but not on other types of mood (e.g., hostility, anger, loss of pleasure), behaviour, or physical changes. Within the treatment-seeking group, Christensen and Oei found no difference between women confirmed with PMS and women not confirmed on number of symptoms, the Beck Depression Inventory, or the State-Trait Anxiety Inventory. In this case the PMS and nonPMS groups did not differ, but the treatment-seeking and non-seeking groups did differ. The results just reviewed demonstrate a distinction between women's assessments of when treatment is required and the diagnostic criteria upon which clinicians make this same decision. The disparity between women's treatment concerns and the criteria used to determine clinical attention is critical in the face of evidence that PMS treatment-seekers may be a high-risk group for suicidal ideation and suicide attempts (Stout, Steege, Blazer, & George, 1986). They have also been found more likely to exhibit dysthymia, phobia, obsession-compulsion, alcohol dependency, and non-alcohol drug dependency (Stout et al., 1986). A direct examination of women's self- labelling with PMS and other definitions of PMS is needed to clarify what women mean by "PMS" when they ask for help. 9 "PMS" as an Illness Label Research on illness cognition suggests that the discrepancies found among labelling, treatment-seeking, and confirmed symptomatology are not unique to PMS. Cameron, Leventhal, and Leventhal (1995) noted that 40% to 65% of visits to physicians are for complaints for which no disease can be detected. Theorists have linked treatment-seeking, physician utilization, and adherence to treatment to individuals' prior beliefs about illness (Baumann, Cameron, Zimmerman, & Leventhal, 1989; Robbins & Kirmayer, 1991). Lau, Bernard, & Hartman (1989) found that illness cognitions accounted for 29% of the variance in propensity to consult a physician. Beliefs about illness have been shown to be more closely related than their actual medical status to how healthy people report feeling (Barsky & Cleary, 1992). This finding is consistent with theories hypothesizing that discrepancies in retrospective and prospective reporting of PMS are attributable to the role of schemata (Koeske & Koeske, 1975; Koeske, 1981, 1986) and of attitudes toward menstruation (Brooks-Gunn & Ruble, 1980). Researchers examining asymptomatic conditions, such as hypertension, have suggested that illness labels are related to how individuals seek information to confirm or disconfirm the illness (Meyer, Leventhal, & Gutman, 1985). How a patient behaves in relation to illness is influenced by past illness experiences, cultural beliefs, and social interactions (Stoller & Forster, 1994). A great deal of variability has been found in the responses of different people to the same symptoms (Stoller & Forster, 1994). Stoller and Forster have interpreted this variability as evidence that individuals do not respond to symptoms but to their perceptions of symptoms. The decision to use an illness label and to seek help is, in part, contingent on the cause to which the patient attributes the illness. Robbins and Kirmayer (1991) 10 demonstrated that psychological help-seeking is greater when symptoms are ascribed to dispositional than to situational factors. The most frequent health care visits are by those who believe their symptoms are caused by a physical disease. There has been little research exploring the use of "PMS" as a label and how its use relates to women's causal attributions of PMS, beliefs about menstruation, and prospective patterns of PMS. This dissertation was an initial exploration of these factors. The Role of Stress in Understanding PMS Koeske (1986) has proposed that the arousability labelling hypothesis necessarily leads to an examination of stress in explaining PMS. Stress may affect level of arousability, increase the salience of negative emotional labels, or increase exposure to negative events. Reid (1985) has suggested that how each woman interprets the severity of her symptoms is likely related to the demands and stresses of her environment. Research into the relationship between life stress and PMS has been expanding, with increasing evidence that they are positively correlated (Ussher, 1992b). Psychological stress has been found to delay menstruation (Parlee, 1973), to suppress ovulation (Abplanalp, Haskett, & Rose, 1980), and to affect both the typicality and length of cycles (Koeske, 1981). Catastrophes and traumas, such as war and nutritional deprivation, have been related to suppressed ovulation and cessation of menses (Sommer, 1978). Sommer (1978) has noted that cessation of menses has been especially associated with strong emotional trauma which is chronic. Social context may directly affect hormonal levels (Birke & Best, 1982). Women have themselves both attributed perimenstrual symptomatology to stress and characterized menstrual symptoms as producers of stress (Woods, Lentz, Mitchell, & Kogan, 1994). Research has supported this interaction by 11 implicating life events (e.g., death of a spouse, job loss) and chronic social stresses (e.g., financial hardship) in the etiology of premenstrual symptoms (Clare, 1985; Futterman, Jones, Miccio-Fonseca, & Quigley, 1992; Kerstner, 1986; Koeske, 1980a; Mitchell, Woods, & Lentz, 1994; Woods et al., 1994; Woods, Mitchell, & Lentz, 1995). In a prospective study by Wilcoxon, Schrader, and Sherif (1976), stressful events accounted for significantly more of the variance in negative mood than did menstrual cycle phase. But conclusions from this study were based on the assumption that the menstrual cycle is an independent variable, the hormonal basis of which is directly responsible for mood changes. Koeske (1981) re-analyzed the Wilcoxon et al. data and demonstrated that normally cycling women were more likely to be in the premenstrual phase at times during the study that coincided with midterm and final exam stress. This cyclical synchrony was not true of women taking oral contraceptives or men. Koeske concluded that stress may have affected menstrual cycle length for these women and that an analysis which assumed that the menstrual cycle is only an independent variable might have been misleading. There is some evidence that daily stressors are more influential than are major life events in PMS reporting (Woods, Most, & Longenecker, 1985). Several researchers have noted that women who are employed outside the home are more likely to report or exhibit PMS symptoms (Coughlin, 1986; Coughlin, 1990; Schnurr, 1988). In one cross-sectional study there was a higher incidence of headache in women doing domestic work in their own homes, and more restlessness and aggressiveness in professional-status women, suggesting that women may report context-specific symptomatology (Huerta-Franco & Malacara, 1993). Prospectively-confirmed PMS has been associated with a lack of autonomy and innovation in the job environment combined with higher work pressure and less personal control (Kuczmierczyk, Labrum, & Johnson, 1992). Access to such personal resources as 12 education and financial safety have been shown to mediate the impact of stressful life events (Mitchell et al., 1994). Brown and Lewis (1993) have attempted to re-focus discussion of stress from stressful life context to "perceived stress" as a variable. Perceived stress is the individual's subjective sense of an event as stressful. Some authors have argued that the evaluation of perceived daily stress provides a more direct and broader prediction of life stress than does measuring major life events (Kanner, Coyne, Schaefer, & Lazarus, 1981). There is some evidence to suggest that the physical experience of stress may not lead to a parallel subjective interpretation of stress. Kirschbaum et al. (Kirschbaum, Schommer, Federenko, Gaab, Neumann, Oellers, Rohleder, Untiedt, Hanker, Pirke, & Hellhammer, 1996) found that male university students who wore estradiol patches for 3 to 4 days did not report subjective stress that matched the levels of physiological stress induced by the Trier Social Stress Test. Although physiological stress responses paralleled endocrine changes, participants did not report subjectively experiencing these changes. The research does provide some support for a link between perceived stress and reports of PMS. Kuczmierczyuk et al. (1992) reported that women showing PMS patterns tended to perceive their family and work environments as more stressful than did women with no PMS. Brown and Lewis (1993) found that women reporting more severe symptomatology also perceived more daily hassles and fewer daily uplifts during the premenstrual than postmenstrual phase. In work by Rubinow and Schmidt (1989), the occurrence of the same events was rated by women prospectively showing PMS as becoming both more frequent and more unpleasant as they moved from the follicular (postmenstrual) to the late luteal phase (premenstrual), by comparison with a nonPMS control group. In a later study by these authors (Schmidt, Grover, Hoban, & Rubinow, 1990), similar results were found. Women showing PMS patterns were not endorsing more distressing events in their lives, but rather were rating the same events as more frequent and troubling. Further support for these findings has come from research by Fontana and Palfai (1994). Women were diagnosed for premenstrual distress (PMD) patterns over 2 months of prospective data. Daily problems and coping strategies were recorded over the same period. Women with PMD more often used catharsis and social support premenstrually than postmenstrually, but there were no phase differences for the control group. This difference became more pronounced for the women with PMD when they perceived stress as being uncontrollable and undesirable. It did not exist for daily stressors that were viewed as controllable or desirable. Fontana and Palfai concluded that perception of stress was state-dependent rather than a reflection of actual environmental stressors per se. They hypothesized that appraisal and coping styles interact over the menstrual cycle for negative stressors. Such an interpretation is consistent with Koeske's schematic framework. Koeske's theory focuses on the interpretation of PMS, rather than on PMS diagnosis. The relationship between actual PMS cyclicity and stress may be of less concern to this theory than the relationship between women's use of the label "PMS" and their actual or perceived stresses. Unfortunately, direct research on the differences in life stress between women who say that they have PMS and those who say that they do not is lacking. Conclusions in this area can only be inferred from research on women's self-reports of premenstrual symptomatology under the induced stress of the Stroop Colour Word Interference Task (Ussher, 1992). Self-diagnosed PMS women reported a higher arousal level during the task than did nonPMS women. But, as previously discussed, symptomatology reports and self-diagnosis are not always consistent. This dissertation examined the relationship 14 between individual women's self-attributions of PMS and their own personal life stress reports. The Role of Relationships in PMS Futterman et al. (1992) found that the primary stressors related to women's self-reports of PMS were interpersonal relationships. Marital satisfaction has been increasingly examined in relation to the menstrual cycle, as both a source of stress and as a dependent variable. Women who live in partnerships with men have been shown to exhibit more problems with PMS than those who are not heterosexually partnered (McDaniel, 1988; Reid, 1985). Research on specific components within heterosexual relationships has demonstrated a link between PMS and lower intimacy (Siegel, 1986), higher marital distress (Siegel, 1987; Stout & Steege, 1985), lower marital satisfaction (Coughlin, 1990; Winter, Ashton, & Moore, 1991), lower sexual relationship satisfaction (Winter et al., 1991), and luteal phase drops in marital adjustment scores (Ryser & Feinauer, 1992). The focus on PMS and romantic relationships has primarily been correlational (e.g., Coughlin, 1990; Rattray, 1986; J. P. Siegel, 1986) or has assumed that the menstrual cycle is an independent variable, in which hormonal changes within the woman affect her role in the relationship (e.g., Hamilton, 1986; Ryser & Feinauer, 1992; Winter et al., 1991). Fewer studies have investigated how the romantic or family environment may affect experiences and perceptions of menstrual cyclicity (e.g., Cohen, 1986). Some have hypothesized that "PMS" is a label that women can use to express anger, dissatisfaction, and gender-violating behaviours in a socially acceptable fashion (Laws, 1985; J . P. Siegel, 1986). Siegel has theorized that emotional outbursts may represent anger and tension which have accumulated 15 throughout the menstrual cycle, but which may be blamed on PMS as a physiological disorder if expressed premenstrually. The woman is consequently excused for her anger as "out-of-character". Koeske (1986) has similarly argued that negative behaviour may be viewed as uncharacteristic of women and therefore as requiring explanation. Koeske and Koeske (1975) have found that internal personality attributions are used to explain moods deemed inappropriate to context, and that situational variables are used to account for moods perceived to be appropriate. In the case of the menstrual cycle, unexpected negative behaviour is more likely to be biologically ascribed. Positive behaviour is viewed as consistent with femininity and tends to be linked to personality and situational causes. McFarlane and Williams (1990) have suggested that women's anger may be considered unacceptable and therefore labelled PMS when it occurs near the expected time of menstruation. Rossi (1980) found that, on average, men who felt "sick" reported feeling angry and unloving to a greater degree than did women. This result suggests that responses to symptomatology may be gender-specific. If role violation is relevant, one would expect women's self-designations of PMS to be related to the context of their lives and their expression of emotions. For example, women might be more apt to say that they have PMS under conditions of higher life stress and an unsafe or unavailable environment in which to express their frustrations. There does appear to be some indirect support for the role violation hypothesis. Stout and Steege (1985) found that 40% of treatment-seekers in their sample exhibited high tendencies to repress or control angry feelings. Moreover, they observed that frequently the content of premenstrual conflict was valid marital issues that were not effectively addressed during other phases. Kuczmierczyk et al. 16 (1992) discovered that direct emotional expression was discouraged by family members of treatment-seekers with prospectively confirmed PMS. Women classified into high-premenstrual-symptom groups, based on a retrospective self-report measure, characterized their family coping strategies as avoidant of self-disclosure, detached socially, and religious or spiritual (Ornitz & Brown, 1993). In the first two studies, however, treatment-seeking and PMS diagnosis unfortunately were confounded. Furthermore, participants were not naive to the menstrual purpose of the investigation. Thus, despite evidence of a relationship to the freedom to express emotion, it is unclear whether this relationship was with women's self-perceptions of premenstrual distress or with their actual premenstrual symptomatology. Because of its use of a retrospective measure and participant awareness of purpose, the Ornitz and Brown study also cannot clarify this issue. There has been no research in which the relationship between women's interpersonal environments and their use of the label "PMS" has been directly examined. Researchers have focused on how relationship variables relate to diagnosis based on researcher criteria, but no one has explored how they are linked to women's own criteria. Without this information, the role violation hypothesis cannot be evaluated clearly. This dissertation examined the differences in relationship satisfaction and freedom to express anger between women who chose to describe themselves as having PMS and women who did not. There is evidence that women who are 30 to 45 years old are more likely to experience premenstrual symptoms than are women between 20 and 30 years (Abplanalp et al., 1980; Golub, 1976; Kramp, 1968). This may be related to physiological changes associated with age. Alternatively, multiple roles may be linked to both stress and menstrually-related changes (McFarlane & Williams, 1990). This dissertation was designed to study an older sample, drawn from the 17 community, rather than the university. Moreover, by studying older women, it was more likely that women with longer-term romantic relationships would be represented. Abuse in Relationships and PMS Winter et al. (1991) have observed that the power dynamics within a relationship may operate interactively with premenstrual symptomatology. This anecdotal observation has not been empirically validated. To more fully address a role violation explanation of women's use of the PMS label, research must expand beyond the study of relationship satisfaction to include an assessment of power dynamics. In particular, abuse within relationships has remained a neglected topic within PMS research. This variable may be especially informative as a measure of safety in assessing the role violation hypothesis. Two studies have shown a relationship between retrospective reports of PMS and sexual abuse history (Friedman, 1982; Miccio-Fonseca, Jones, & Futterman, 1990) and a third study has demonstrated a correlation between sexual trauma and prospectively confirmed PMS (Paddison, Hartley, Lebovits, Strain, Cirasole, & Levine, 1990). Unfortunately, participants in these studies were not naive to the menstrual interest of the researchers, which may have resulted in inflated symptom reports (AuBuchon & Calhoun, 1985), making interpretation of the results difficult. Miccio-Fonseca et al. (1990) did find evidence that sexually traumatized women indicated experiencing greater life stress and tended to keep their sexual abuses secret, presenting for therapy based on other difficulties. "PMS" may be one form of presentation used to seek treatment. History of psychiatric hospitalization in general has been found to be higher among women seeking treatment for PMS (Miccio-Fonseca et al., 1990; Paddison et al., 1990). 18 Sexual abuse history has been linked to low self-esteem, depression, marital difficulties, greater incidence of family psychiatric histories, and problems in forming successful intimate relationships (Miccio-Fonseca et al., 1990). Both current and past patterns of abuse in a woman's life may therefore operate as chronic, daily stressors. Researchers have not investigated how emotional and physical forms of abuse relate to PMS. Moreover, there has been no research explicitly addressing how women's experiences with abuse relate to each of their prospective menstrual patterns and their self-ascriptions of PMS. If the role violation hypothesis is valid, women reporting current and past abusive relationships would be more likely to say that they have PMS regardless of whether their prospective data confirmed PMS. One purpose of this dissertation was to compare the reports of women who self-designate as having PMS and women who do not with regard to their childhood and adulthood histories of physical, emotional, and sexual abuse. In addition, ongoing psychological abuse by heterosexual women's romantic partners was measured (using the Psychological Maltreatment of Women Scale) and related to self-designations of PMS. This was the first PMS study to explore non-physical forms of abuse history, the first to relate abuse to women's own self-diagnoses, and the first to study any form of abuse history among participants who were naive to the menstrual purpose of the research. PMS and Depression The most commonly reported adult symptom of childhood sexual abuse is depression (Paddison et al., 1990). During the past decade, researchers have increasingly focused on the similarities of symptom patterns between non-endogenous depression and PMS (Halbreich et al., 1983; Rosen, Moghadam, & Endicott, 1988; Siegel, 1987). Some researchers have suggested that PMS may be a phasic exacerbation of already existing mood, behavioural, and physical 19 symptoms (e.g., Clare, 1985; Jorgensen, Rossignol, & Bonnlander, 1993). Because of symptom similarity and possible co-morbidity, screening participants for depression has become a commonly recommended research practice, to ensure that diagnosis represents "pure" PMS (Cumming et al., 1991; Siegel, 1987). Data collected across all phases of the menstrual cycle have been used to separate cyclically-related PMS from non-PMS dysphoria occurring throughout the cycle (Chisholm, Jung, Cumming, & Cumming, 1990; Christensen, Board, & Oei, 1992). Because one purpose of this dissertation was to examine abuse history, the procedure of screening potentially depressed women from the sample was not used. Moreover, had these women been screened from the study, participants reporting higher levels of life and chronic stress might also be excluded. In both cases, the range of important stress variables may be limited. For the purposes of this thesis, the women in this research project who were excluded from previous analyses of mood data, based on Beck Depression Inventory (BDI) scores exceeding 15 (McFarlane & Williams, 1994), were included. One caution is that the women who have higher BDI scores and report that they have PMS may be reporting actual patterns in which depressive symptoms are exacerbated premenstrually. A second issue is whether women who say that they have PMS and demonstrate higher BDI scores are over-reporters of symptomatology in general. Both of these issues are explored in discussing the results. Variability Among Women Research using a singular concept of PMS has been criticized for combining diverse behavioural, physical, and psychological symptoms without clear specification of what is being measured (Christensen & Oei, 1995; Halbreich & Endicott, 1985b). All symptoms may be weighed equally without regard to the experiences of the woman or symptomatic origin (Abplanalp, 1983). The number of 20 symptoms endorsed has often been equated with severity, rather than measuring severity of individual symptoms or types of symptoms (Busch, Costa, Whitehead, & Heller, 1988). It may be that groupings of similar symptom types (e.g., affective symptoms) interact differently with the context of women's lives and that these individualized effects are obscured by simply classifying women dichotomously as having or not having PMS (Halbreich et al., 1983). High interindividual and intraindividual variability has been shown in menstrual cycle changes. Symptoms, phase length, and intensity of flow have been found to vary from cycle to cycle and from woman to woman (Abplanalp, 1983; Ekholm & Backstrom, 1992; Rodin, 1992; Van der Ploeg & Lodder, 1993). Women have also demonstrated a great deal of variability in their own definitions of PMS (Birke & Best, 1982). Parlee (1978) has cautioned that using statistically averaged data at the group level may obscure variability and result in a loss of information. In a factor analysis of the Menstrual Distress Questionnaire (MDQ), Woods, Most, and Dery (1982) concluded that the use of group analyses notably obscured the individual variability of reports. In their previous study of the sample used for this thesis, McFarlane and Williams (1994) similarly found that use of parametric techniques such as analysis of variance on group (normative) data masked the extensive individual patterns of cyclicity that were revealed in idiographic analyses. Koeske (1981) has argued that menstrual cycle researchers are often forced to rely on data collection and analytic techniques that are not designed to represent complex and dynamic systems of variables. She noted that analysis of variance relies on a clear demarcation between independent and dependent variables, the representation of one variable with limited values, and averaged group data. These requirements may not be appropriate for an interactive, individual model. She has suggested 21 instead that sufficient baseline data be collected to allow variable changes to be compared to the person's own norm and expressed as a deviation. Schnurr (1988) followed this proposal in the use of effect size to measure premenstrual change within the context of the woman's own variability. However, Schnurr based this calculation on a premenstrual-postmenstrual difference score rather than on data from the entire menstrual cycle. Hence, the woman's variability was assessed against a single other cycle phase that was assumed to be normative, rather than against her variability across phases. McFarlane and Williams (1994) used each individual woman's own mean and standard deviation for each of seven mood variables to calculate uncharacteristic change. Prediction analyses were used to determine whether nonaverage experiences occurred statistically more often for particular menstrual phases, lunar phases, or days of the week for each variable. Clusters were created using those variables that were found statistically significant for each woman and the final clusters were examined for cyclical patterns. This technique revealed individual variability in the composition of symptom clusters and in the definition of "uncharacteristic change". The McFarlane and Williams (1994) study focused specifically on cyclicity in mood variables. No relationship between women's prospective menstrual patterns and their reports of having PMS was found. But perhaps those women used different symptoms to define their cyclical experiences. This dissertation was a further, idiographically-based analysis of 15 variables for which daily data were collected to see which factors demonstrating cyclicity best reflect women's self-labelling. In addition, women who said that they had experienced PMS were directly asked how they were defining it and this definition was compared to their life contexts. Finally, the context of stressors in these women's lives was discussed in 22 relation to their own assessments of what has caused their personal PMS experiences. Purposes of this Study This dissertation was designed to explore the discrepancy between women's self-designated PMS and prospectively-based researcher-designated PMS. This was one of the first studies to systematically investigate factors related to women's use of the label "PMS". The first purpose of this study was to examine the nature of prospective cyclical change in a wide variety of variables for women who report having and not having PMS. This information was collected to explore whether women place the same emphasis on affective changes as have McFarlane and Williams (1994) and the DSM-IV PDD diagnosis. Patterns of menstrual cyclicity for 15 variables categorized into 4 different symptom types were analyzed to investigate whether a particular type of symptom was more salient for women labelling themselves with PMS. In addition, women's self-designations regarding PMS were compared first to researcher-designations of PMS based on the criteria developed by McFarlane and Williams (1994), and then to PDD, based on the criteria of the DSM-IV (APA, 1994). If the explanation for the discrepancy between self-labelling and prospective confirmation found by McFarlane and Williams (1994) is that women do not base their labels on mood changes, then the prospective measurement of a more diverse set of variables might result in a better match. The second purpose of this dissertation was to investigate whether women who said that they had PMS were mislabelling menstrual cyclicity patterns occurring in a non-premenstrual phase. McFarlane and Williams (1994) found some evidence for patterns of non-premenstrual negative cyclicity and for patterns of non-23 premenstrual positive cyclicity in women who said that they had PMS and did not meet researcher criteria. This dissertation further explored this explanation. Research has shown a link between illness labelling and schematic representations of illness (Baumann et al., 1989; Meyer et al., 1985; Robbins & Kirmayer, 1991; Stoller & Forster, 1994), but there has been little study of women's use of "PMS" as a label. A third intent of this dissertation was to explore what attitudes and definitions women are describing when they label themselves as having PMS. In addition, it explored the role that possible situational cues, such as perceived stress, relationships, and abuse, might play in self-designation. The fourth purpose was to study the relationship of abuse history to women's beliefs about the origin of PMS and their self-designations of PMS. In particular, this study expanded upon previous work by obtaining information about emotional, physical, and sexual abuse in both adulthood and in childhood. In addition, psychological maltreatment by the woman's current heterosexual partner as well as partner abuse history was assessed. To further assess whether women reporting abuse were using the PMS label to be able to safely express dissatisfaction, their reports of anger arousal, the amount of anger that they experience at home, and the amount of sadness that they experience at home were measured. This study attempted to shed light on the role violation hypothesis by examining whether women who feel they are in unsafe environments or poor relationships and who report rarely expressing anger are more likely to describe themselves as having PMS. Finally, because the measures used in this study were self-report, it could address the question of whether women who label themselves as having PMS tend to report more symptomatology and dissatisfaction generally. 24 Method Participants Participants were recruited through Vancouver city newspapers and posters for a study of emotional, behavioural, and health patterns. Posters were placed at community centres, the university campus, in local businesses, and in various public locations. A total of 418 individuals responded by telephoning (295 women, 99 men, 24 of unknown gender) and 110 completed the study. Data from some of the female participants were excluded for this dissertation because of cycle length (only those cycling between 21 and 35 days were included), hysterectomies or ovariectomies (11 excluded), pregnancy, and chronic illnesses (5 excluded). Only one woman identified as a lesbian, so analyses of group differences based on sexuality were not possible, and this dissertation focussed only on heterosexual women. In addition, the data collected from 35 men, 20 to 79 years (mean = 40.8 years), were not relevant to the primary purpose of this study. The data from 58 women, aged 20 to 48 years (mean = 34.5 years), who supplied complete prospective data were analyzed. Their romantic and sexual relationship status, self-described ethnicities, race, religious affiliations, occupation, employment status, and educational backgrounds are presented in Table 1. The occupational categories are based on a system presented by Strieker (1988), with "mother" and "artist" added for this study. Participants were unpaid volunteers but were entered into a random draw for one $250.00 prize at study completion. Of the 58 women in this study, 9 were using oral contraception during data collection. Five of these women said that they had PMS but their prospective data did not meet researcher-criteria of PMS or PDD. Four women said that they had never had PMS, one of whom met researcher criteria for PDD, but not PMS. Previous research has indicated few consistent menstrually-related psychological or behavioural differences between oral-contraceptive (OC) and normally-cycling (NC) Table 1 Demographics of sample Romantic Relationship Status Married or cohabiting (mean=7.9 years) 47.7% In a romantic,non-cohabitating relationship,sexual 4.6% In a romantic,non-cohabitating relationship,celibate 1.5% Uninvolved romantically,sexually active 15.4% Uninvolved romantically.celibate 18.5% Divorced or separated 12.3% Identified as heterosexual 100.0% Sexually Active Sexually active with another Not sexually active with another .70.8% .29.2% Race Women of Colour 7.7% White women 92.3% Self-described Ethnicity Afro-Brazilian 1 Anglo 1 Austrian 1 British 7 Chinese 4 Danish 1 English 12 Estonian 1 European 1 French 2 German 6 Icelandic 1 Irish 2 Italian 4 Participant did not identify 11 Self-described Religion Atheism, Agnostic, n.a. . 30.8 % Buddhism 3.1 % Christianity 43.1 % Science of Mind Church .. 1.5 % Jewish 3 Latina Peruvian Mestizo Polish Polynesian Russian Scottish Slovak Slovic South African Spanish Swedish Ukranian 2 Welsh 1 Participant did not know . 3 Judaism 3.1 % Personal 16.9 % Self-described Employment Status Employed full-time 50.8% Employed part-time 29.2% Unemployed 13.8% FT student 7.7% PT student 23.1% 26 Highest Level of Education Obtained Completion of elementary school 0.0% Completion of secondary school 9.2% Post-secondary coursework or upgrading 16.9% Post-secondary degree or diploma work 18.5% Completion of post-secondary diploma or degree 44.6% Post-graduate work 0.0% Completion of post-graduate degree 9.2% Unknown 3.1% Occupation (Strieker, 1988) Professional Worker 29.2 % Clerical or Administrative Support Worker 21.5 % Executive or Administrative Worker 7.7 % Student 7.7 % Sales Worker 6.2 % Technical Worker 4.6 % Personal Service Worker 4.6 % Craft or Precision Worker 3.1 % Homemaker/Personal Home Worker 3.1 % Transportation Worker 3.1 % Unemployed 3.1 % Handler, Helper, or Labourer 1.5 % I Don't Know/In Transition 1.5 % Machine Operator or Assembler 1.5 % Mother 1.5 % Artist 1.5% 27 women (Elliott, 1992; McFarlane et al., 1988; Stephenson, Denney, &Aberger, 1983), suggesting that combining these groups might be appropriate. Moreover, previous analyses conducted on this sample for mood variables showed no significant differences between OC and NC women for prospective PMS designations or self-designations (McFarlane & Williams, 1994). Some researchers have, however, found differences in physical and emotional symptomatology between O C and NC women (AuBuchon & Calhoun, 1985; Koeske, 1981; Wilcoxon et al., 1976; Woods, Most, & Dery, 1982). Procedure Each woman participated in three phases of data collection: an initial interview, a daily questionnaire, and a final interview. Initial interview. Participants began the study with a 1-hour individual interview session. During this session, each individual completed a set of questionnaires (the initial interview package) which included menstrual (e.g., discomfort), lifestyle (e.g., exercise), and demographic (e.g., age) questions, as well as the Beck Depression Inventory (1967) (see Appendix D). Questionnaire responses were subsequently clarified by the interviewer and the study procedures were thoroughly explained. The initial interview package included all of the questions required for the screening criteria. Daily questionnaire. Participants then completed the same daily chart each day for 12 to 18 weeks (mean = 15.6 weeks). The data from all days for which each participant completed daily charts were used. Thus, the number of full cycles varied from 3 to 5, and partial cycles were included in the relevant phases (e.g., premenstrual, midcycle). The daily chart was a two-sided, 8 1/2" x 14" sheet, with 9-point bipolar scales for 24 items. This dissertation focussed on 15 of these variables: amount of exercise, quality of sleep, body experiences, eating pattern, 28 ability to concentrate, patient/impatient behaviour with others, paid work performance, domestic work performance, time spent on leisure, quality of personal relationships, quality of professional relationships, mood pleasantness, mood stability/variability, anger, and hurt feelings. Because the statistical procedures used for this dissertation required that each variable have "positive" and "negative" anchors, those items for which it was difficult to determine these poles were omitted (e.g., it was not clear whether having a great deal of responsibility at work, having no responsibility at work, or a middle rating should be interpreted as most positive across participants). A question asking whether the woman was menstruating was embedded among other health questions to conceal the menstrual interest of the researchers. The participant noted the date, time of day, and whether the chart was being completed on time at the top of each form. Participants were encouraged to complete the chart at a similar time each day. They were instructed that the chart for a forgotten day should be filled out as soon as remembered, but that any items that could not be recalled clearly should be left blank. For all close-ended questions, women were asked to choose only the one alternative which best represented the day being reported. They were specifically instructed to evaluate each item according to their own standards rather than the standards of others with whom they might relate. The completed daily forms were mailed to the researchers weekly by the participants to allow continuous data entry, to ensure safe and confidential storage of data, and to avoid retrospective revisions to the charts by participants. Researchers telephoned each woman bi-weekly to answer questions and to encourage continued participation. The daily chart included some questions about health and physical pain, uncharacteristic-for-me experiences, and importance ratings for each of the 24 bipolar scales. These items were interpreted with great variability by the 29 participants and were not analyzed for this dissertation. The complete questionnaire and the written instructions received by each participant are in Appendix E. Final interview. Upon completion of the daily data, each participant came in for a final session that included a semi-structured interview, various questionnaires, and feedback about her or his own prospective data profiles (see Appendix F for a copy of the complete final interview package). The full purpose of the study was not revealed to any participant until all individuals had either provided all data or dropped out (79 women dropped out after the initial interview; 83 dropped out during daily data collection). None of the participants successfully guessed the purpose when asked at the end of the study. Measures from the final interview that were analyzed for this dissertation include the Multidimensional Anger Inventory (MAI) (Siegel, 1986), the Marital Need Satisfaction Inventory (Stinnett, Collins, & Montgomery, 1970), the Menstrual Attitude Questionnaire (MAQ) (Brooks-Gunn & Ruble, 1980), the Psychological Maltreatment of Women Inventory (PMW) (Tolman, 1989), and the Daily Hassles and Uplifts Scale (DeLongis, Coyne, Dakof, Folkman, & Lazarus, 1982). Four of these measures provide subscale scores. The MAI measures hostile outlook, the individual's relative tendency to express anger outward or inward (mode of anger expression), the range of situations that tend to elicit anger, and the ease with which anger is aroused (anger-arousal). The MAQ assesses five dimensions of menstrual attitudes, including the extent to which the woman views menstruation as predictive, debilitating, bothersome, natural, and the extent to which she claims that menstruation should have no effect. 3 The Psychological Maltreatment of Women Inventory allows the woman to rate her 3 The subscale of the Menstrual Attitude Questionnaire developed by Brooks-Gunn and Ruble (1980) to measure women's "denial of the effects of menstruation" is referred to in this dissertation as "menstruation should be ignored", to avoid confusion with other clinical (e.g., psychodynamic) uses of the term "denial". 30 partner's treatment of her along two factors labelled domination-isolation and emotional-verbal. The Daily Hassles and Uplifts Scale yields frequency and intensity scores for each of hassles and uplifts. The frequency and intensity scores for participants were highly correlated (r for hassles=.97, r for uplifts=96). Mean intensity and mean frequency scores were multiplied for each of hassles and uplifts and these products were used for analyses (DeLongis, personal communication, Jan. 10, 1997). (See Appendix G for reliability and validity information on these scales.) Other scales completed during the final interview session that were excluded from analyses were created for the longitudinal project, but in a separate study subsequently were found to have poor reliability (See Appendix G). One series of retrospective questions about menstrual, lunar, and day of week patterns for the mood variables was not related to the purpose of this dissertation and therefore not used. Participants answered questions about their history in adulthood and in childhood of emotional, physical, and sexual abuse, as well as their history of emotional, physical, and sexual abuse by a romantic partner. They also rated the current frequency and intensity of their feelings of anger and sadness at home. Frequency and intensity of anger at home were combined into a single scale, "Amount of Anger Experienced at Home" and frequency and intensity of sadness were combined into a single scale, "Amount of Sadness Experienced at Home". As preparation for and part of this dissertation, I conducted a test-retest reliability study on a separate undergraduate sample of 199 women and 218 men on the questions which had been developed for this research project. Good test-retest reliability was found for the abuse, sadness, and anger questions analyzed for this dissertation. Internal consistency was good (Cronbach's alpha=.81) for the Amount of Sadness 31 Experienced at Home scale and acceptable (Cronbach's alpha=. 68) for the Anger Experienced at Home scale (see Appendix G for a summary of the reliability study.) In a questionnaire completed as part of the final interview package, each woman was asked whether she had ever had PMS and if so, what she believed to be the cause of her PMS. In addition, she was asked to indicate whether she was employed and whether she was in a romantic relationship at the time of her PMS. Researchers Data collection was conducted by a research team of 18 women and 2 men who participated at various periods throughout the longitudinal project. Interviews were conducted by three senior undergraduate and two graduate students who were in clinical, social work, or counselling training. These interviewers, all of whom were white women, varied in age and socioeconomic background. Telephone contact calls were made by several senior undergraduate women who varied in socioeconomic background and ethnicity. If a crisis or abuse situation was reported in the daily charts, the participant was immediately contacted by a graduate student interviewer. Undergraduate and graduate researchers participated in data coding, entry, and analysis. Clerical contributions were provided by the departmental secretary at UBC. Most of the researchers were heterosexual and were in varied romantic and nonromantic relationships at the time of data collection. Self-designation of "PMS" Two conditions were required for classification as "Current-Say" (women who say they currently have PMS): (a) an affirmative response to the question, "Was there ever a time when you think you had PMS?" and (b) a response to "When?" or "How long did the PMS last?" that confirmed that the participant believed she currently experienced PMS. If a woman answered that she had never had PMS, 32 she was classified as "Never-Say" (women who say that they have never had PMS). Of the total sample, 35 women reported currently having PMS and 23 women did not. There were no significant differences between Current-Say and Never-Say women for age (Current-Say M=35.09 years, range 22 to 48 years; Never-Say M=33.43 years, range=20 to 45 years), reports of average length of menses, number of children, whether the woman had ever been pregnant, cycle length, education, duration of current romantic relationships, employment, reports of health status (good or not good), room-mate or cohabiting status, marital status, regularity of menstruation, and status as a student. The two groups differed on only two demographic variables. Women's self-designations of PMS were significantly related to their employment status,x2=4.80, rj<.05. Of the 41 women who were employed, 60% of Current-Say women worked part-time whereas 25% of Never-Say women worked part-time. Second, women who said that they currently had PMS were more likely to report being (heterosexually) currently sexually active, phi=.31. rj<05. About half (54%) the Never-Say women reported being heterosexually active whereas 83% of the Current-Say women did so. The primary interest of this study was in comparing the Current-Say and Never-Say groups. Participants' responses indicated, however, that women who said that they had PMS varied in how consistently they believed it to occur. The analyses were therefore repeated using 3 groups: Always-Say (women who said that they had always had PMS), Sometimes-Say (women who said that they had had PMS intermittently), and Never-Say (women who said that they had never had PMS). Many of these analyses yielded the same results as the comparisons between the two Say groups. Findings that were unique have been reported. 33 Coding of Women's Causal Explanations for PMS Procedure. The women who said that they had PMS also answered an open-ended question about what they believed to cause their own PMS. These responses were subsequently coded using a system developed by four members of the research team. They were white women of varied ages and socioeconomic backgrounds. To develop the coding system, all responses were converted to transcripts and labelled only by participant identification numbers. Each researcher independently developed as many categories and subcategories as she found necessary to describe the varied responses. The final coding system was created through a process of comparing and negotiating these categories. Since the participants interpreted and answered the question in different ways, the resulting categories reflected several types of information. For example, many women did not restrict their responses to causal theories, but also included ideas about the character and consequences of PMS. Responses that represented notions of character and consequences rather than cause were omitted from the analyses for this dissertation. An effort was made to minimize researcher interpretation of the participant's response. Category creation emphasized a parsimonious reading of the response rather than implied or intended meanings. Unsupported assumptions about what a participant might be trying to express were avoided by focussing on the literal coding of statements. Transcripts were then independently coded by two researchers to permit calculation of interrater reliability. Although the woman's response as a whole was considered in these coding decisions, each coder recorded the statements she found most salient to her decision under the coding category selected. This procedure was used to facilitate the negotiation of discrepancies. Ageement was obtained for 81% of responses, based on the two 34 coders' independent categorizations. All but one of the disagreements occurred between the cause-dv and cause-both categories. Disagreements were resolved by discussion between the coders. Coding system. Causal attributions included statements by the participants which represented notions of the origin of PMS in their own lives or for women in general. Responses were coded into one of four subcategories: cause-iv, cause-dv, PMS as mystery, cause-both, or no response. A Cause-iv response was defined as any cause in which the menstrual cycle was characterized as an independent variable. The origin of PMS was attributed to physiological or biological origins or explicit reference was made to PMS as inevitable and natural to women's reproductive cycles. This type of response was representative of a biomedical model of PMS. To meet coding criteria for this subcategory, the whole response had to indicate a purely physical explanation not combined with or mediated by the social environment. The second subcategory, cause-dv. included responses that characterized the menstrual cycle as a dependent variable. PMS was attributed solely to socio-environmental origins with no simultaneous or more basic biological origin. If an answer combined physiological and socio-environmental causes in a mutually operating manner it was coded as cause-both. PMS as mystery was used to list any answer that characterized PMS as a phenomenon the woman felt was unexplainable either by her, by professionals, or by both. This subcategory was separated from women who chose not to respond, who were coded under no response. Since a lack of response is difficult to interpret, it was treated as missing data in the statistical analyses. Definitions of Menstrual Phases Each cycle of each woman was coded into five menstrual phases: menstrual, postmenstrual, midcycle, post-midcycle, premenstrual. The phases were 35 calculated based on the length of each cycle for each individual. The menstrual phase was designated as the days of menstrual bleeding reported by the participant on her daily charts. Consistent with Frank's (1931) original definition of PMT, the DSM-IV PDD criteria 4, and the Moos Menstrual Distress Questionnaire, the premenstrual phase was defined as the 7 days prior to menses (see Appendix A for a further discussion of these definitions). The midcycle was calculated as the middle five days of the cycle. The remaining days were equally divided into the postmenstrual and post-midcycle phases. A question about breast tenderness was included in the daily charts to help determine probable ovulatory cycles, but it was not possible to train participants to properly palpate their breasts premenstrually, as it would have alerted them to the menstrual interest of the study. Thus, the terminology chosen to describe the menstrual phases emphasizes the timing of the phase in the cycle rather than hormonal characteristics (see also McFarlane & Williams, 1994 for a discussion of this methodological decision). Analysis of Daily Variables Idioqraphic analyses. The mean and standard deviation of each of the 15 daily variables was calculated across all days in the study for each individual. These two measures were used to indicate each woman's average or "normal" range for each variable. Any day rated higher than one standard deviation above her mean for a particular variable was considered to be "high" for her for that variable; any day rated as more than one standard deviation below her mean for that variable was considered "low" for her for that variable. All variables were coded during computer analysis so that "highs" were "positives" and "lows" were "negatives". Prediction analyses were then applied to determine the statistically "The DSM-IV defines the premenstrual phase as the week prior to menses and the first few days of menses. For this project, days on which women reported menstrual bleeding were excluded form the definition of the premenstrual phase. 36 significant occurrence of "positives" and "negatives" for each menstrual phase. The manual developed by McFarlane and Williams in which this prediction analysis procedure is described is in Appendix H. The criteria for researcher designations of PMS and PDD. Each woman's menstrual cycles were examined for evidence of a PMS pattern using all 15 variables. The women were assigned to researcher-designated "PMS" or "nonPMS" and "PDD" or "nonPDD" designations. Two sets of criteria were used to determine whether each woman's prospective data showed evidence of a PMS pattern. The first set of criteria was based on those used by McFarlane and Williams (1994). The women were assessed individually. Each menstrual phase was examined for the presence of a "terrific" (T) or "miserable" (M) cluster. The criteria for a T were (a) the presence of at least three positive variables that attained statistical significance for the participant in the prediction analysis procedure, and (b) that at least 75% of the significant variables within that phase were positive. The parallel criteria for a M cluster were (a) the presence of at least three statistically significant negative variables, and (b) that at least 75% of the significant variables within that phase were negative. The rationale proposed by McFarlane and Williams for the use of a cluster of three variables to represent a menstrual phase pattern was that three or more variables would create both a more salient impact for a woman and a more reliable unit of analysis than would single variables. A "PMS" classification was assigned if: (a) a M cluster occurred in the premenstrual phase, and (b) no other phase of the menstrual cycle was characterized by a M cluster. If a woman did not meet both of these stipulations, she was designated as "nonPMS". The second set of criteria used a 5-item cluster based on the diagnostic standards of Premenstrual Dysphoric Disorder (PDD) defined by the DSM-IV (See Appendix C). Both T and M clusters were defined so that relative positive and 37 negative cyclicity could be analyzed. To more closely match the DSM-IV PDD criteria, women were assigned a "PDD" classification if: (a) a 5-item M occurred premenstrually, (b) no M was present during the postmenstrual, midcycle, or post-midcycle phase 5 , and (c) at least one of the five variables comprising the premenstrual M was a mood variable. The DSM-IV PDD diagnosis does not require a symptom-free menstrual phase, so a menstrual phase M did not preclude a "PDD" classsification. If a woman did not meet all of these criteria, she was designated as "nonPDD". One criterion of the DSM-IV PDD diagnostic category is that premenstrual difficulties should markedly interfere with the patient's usual work or school or social activities. All of the women classified as PDD had at least one significantly negative item that was either a relationship or a concentration and work performance variable. These designations were used as a tool to determine the match between potentially diagnosable PMS or PDD patterns and women's self-designations. The small number of women who met researcher criteria precluded a reliable interactive analysis of researcher-designations by self-designations in this study. Analysis of Single-measurement Variables Some methodological errors to consider. As a result of copying errors, two scales required conditional use and interpretation. The omission of items from the Marital Need Satisfaction Scale (MNS Scale) and the Multidimensional Anger Inventory (MAI) were realized only subsequent to administration and coding. The Marital Need Satisfaction Scale was developed as a 24-item scale (Stinnett, Collins, & Montgomery, 1970). Only the first 15 items were presented in the final interview 5 This is a slightly stricter criterion than the DSM-IV which specifies a symptom-free postmenstrual phase. Although the midcycle and post-midcycle phases should be symptom-free to meet PDD standards, the DSM-IV also notes that ovulatory symptoms may be present for some women. No such cases occurred for this sample. 38 package of this study. These questions were considered informative to this dissertation and were therefore evaluated as a potential scale for analysis. Cronbach's alpha for these 15 items for this sample was .92, suggesting high internal consistency. The 15 items were therefore summed into a total scale score for each participant. The Multidimensional Anger Inventory is a 38-item 4-factor questionnaire. Overall internal consistency was found to exceed .8 in the original scale development. The first 32 items were used in this study and Cronbach's alpha indicated high internal consistency (.86) for this sample. The missing 6 items were exclusively those which had loaded on the "range of anger-eliciting situations" and "hostile outlook" factors in the original study (Siegel, 1986). These factors were therefore elimiinated from analyses and the first two factors, "anger-arousal" and "mode of anger expression", were retained. Unfortunately, the acceptable internal consistencies for the mode of anger dimensions found by J. M. Siegel were not replicated in this sample (anger-in=.44), and therefore this subscale was not used for this dissertation. The anger-arousal subscale did demonstrate high internal consistency (.87) for this sample and was therefore used for the analyses. The relationship between the content represented by the MNS Scale scores, the MAI subscale score, and women's decisions to label themselves with PMS was of interest in this study. Because of item omission, overall scores on these "scales" cannot be compared directly either to the original normative sample scores or to the scores reported in other studies. Our purpose, however, was neither diagnostic nor psychometric in relation to these scales. It was the relative scores of Current-Say and Never-Say women, not the absolute scale scores which were important for this dissertation. With their high internal consistency, the partial MNS Scale and the complete anger-arousal subscale of the MAI seemed to be appropriate measures to 39 help inform the question of how women's feelings about their lives relate to their self-designations. 40 Results Approach to Normative Data Analysis Several types of statistical analyses were considered for this dissertation. Given the longitudinal nature of the data, it was intuitively inviting to consider a time series approach. One possible technique is the Hierarchical Linear Model (HLM). It provides a powerful measure of between-subject and within-subject change (Bryk & Raudenbush, 1987). It is particularly useful because the number and timing of observations may differ across participants (Bryk & Raudenbush, 1987). It would be an ideal approach, for example, for studying how some of the single-measurement variables predict menstrual cycle patterns. The purpose of this dissertation was not to study variables that may affect menstrual cyclicity. Rather, it addressed the differences between women who say that they currently have PMS and women who say that they have never had PMS. Since this variable, that is, the "say" variable was only measured at one point, it does not provide the repeated measures dependent variable data required for HLM analysis (Bryk & Raudenbush, 1987). (For a further discussion of time series techniques, such as Fourier and spectral analysis, see Appendix H.). A second statistic considered was discriminant function analysis (DFA), used in a descriptive approach (Huberty, 1975). Given the nature of this study and the goal to assess the role of various factors in relation to each of the two say groups, DFA does at first appear to be a logical choice for statistical analysis. The primary objective in using DFA is to identify characteristics that optimally distinguish two or more mutually exclusive, pre-defined groups (Cocozzelli, 1988; Dattalo, 1994; Franzen & Golden, 1985; Gondek, 1981). DFA, however, requires a substantial sample size for reliable results. Dattalo (1994) characterizes sample sizes of less than 400 as small and cautions against assumption violation for such studies. 41 Whereas some researchers recommend that the ratio of sample size to number of predictors should be 10:1 (Franzen & Golden, 1985), others have recommended a ratio of 20:1 (Stevens, 1996). Smaller sample sizes are prone to shrinkage on cross-validation and must be interpreted with extreme caution (Barcikowski & Stevens, 1975; Borgen & Seling, 1978; Fletcher, Rice, & Ray, 1978; Huberty, 1975; Tate, 1983). In Monte Carlo results generated by Fletcher et al. (1978), a sample size approximating the one in this dissertation and a similar sample to variable ratio produced 35% shrinkage in external cross-validation. Another drawback of DFA for this dissertation is that it cannot evaluate how the say variable interacts with the menstrual phase in relation to changes observed for the daily variables, an important aspect of this study. A jackknife procedure has been suggested for smaller samples (Stevens, 1996), but it is limited to a context in which predictors have been selected with a strong theoretical rationale based on previous findings (Franzen & Golden, 1985). Logistic regression has been used instead of DFA in cases of possible or suspected assumption violation (Dattalo, 1994; Stevens, 1996). But Dattalo (1994) cautions that this statistic requires at least 50 cases per parameter and that little is known about how it performs under various conditions. (For a discussion of why other techniques, such as cluster analysis and factor analysis, were not considered, see Appendix A.). Hypothesis-testing about significant group differences may be done with MANOVA (Fletcher et al., 1978). In contrast to DFA, MANOVA is a powerful statistical technique that is robust to assumption violation, even in the case of small sample sizes or unequal groups (Glass & Hopkins, 1996; Norusis, 1985; Stevens, 1996). Pillai's Trace, in particular, provides a reasonably accurate significance level even when assumptions are violated (Norusis, 1985), so it was used in this 42 dissertation for all the MANOVA analyses. A statistical technique that compares means (e.g., MANOVA, ANOVA, t-test) requires a much smaller sample size than that required for correlational techniques (e.g., DFA, factor analysis, regression). MANOVA is therefore more suitable than DFA to test for group differences in this dissertation, with a sample size of 58 and unequal group sizes. Stevens (1996) has suggested that MANOVA and univariate followups are preferable to DFA if the researcher is more interested in how specific variables relate to group differences than in linear combinations of variables. In a Monte Carlo study by Hummel and Sligo (1971), the use of followup univariate t-tests to MANOVA (p_<05), each conducted at p_<05, successfully kept the overall alpha-level under control for cases in which the multivariate null hypothesis was true. Stevens (1996) notes that this procedure is an important one for obtaining sufficient power to detect differences in a small or moderate sample size. In this dissertation, univariate F tests (p_< 05) were used for followup analyses where MANOVA indicated significance at p_<05. For multiple comparisons conducted, the Tukey method was used to provide a family-based error rate (p_<.05) and a relatively conservative test of effects (Glass & Hopkins, 1996). Stevens (1996) has recommended that to ensure greater statistical power and to protect optimally against Type I error, a larger number of variables (e.g., more than 10) should be divided up for analysis rather than used in one analysis. For both this reason and conceptual reasons, the single-measurement variables in this dissertation were divided into five MANOVAs. This division was based on: (1) conceptual similarity: variables were grouped into the same analysis if they formed meaningful conceptual units (e.g., Beck Depression Inventory, sadness at home, anger at home, anger arousal) (2) missing data: the Marital Need Satisfaction Scale (MNS) and the Psychological Maltreatment of Women Inventory 43 (PMWI) were each analyzed separately in order to maximize the sample size for each analysis. Not all participants were in romantic relationships and MANOVA omits cases that contain missing data. A t-test was then used to analyze whether the two say groups significantly differed in mean MNS scores and a MANOVA was used to analyze the two subscale scores of the PMWI. (3) measures containing subscales: subscales of a single measure were analyzed together. The Statistical Package for the Social Sciences (SPSS) programme for Windows was used for all statistical analyses. MANOVA considers variables as jointly dependent and considers the intercorrelations between variables (Norusis, 1985). But when univariate followups are used, ANOVA ignores the interrelationships among variables. In this dissertation, the variables were grouped into several MANOVAs. Thus, some correlated variables were analyzed separately and their intercorrelations were not considered. Correlations among all single-measurement variables are provided in Table 2. Of those 253 correlations, 13 would be expected to be statistically significant by chance (a=.05). Most of the 38 significant correlations occurred between conceptually related measures and/or related subscales. Meehl (1990) has termed the average pairwise correlation among variables the "crud factor". The crud factor is symptomatic of the reality that in the social sciences, "everything is correlated to everything" (Meehl, 1990, p. 210). Meehl (1990) notes that alpha is an unrealistic estimate because the proper baseline for hypothesis testing is not H Q , but the crud factor. The mean of the absolute value of the pairwise correlations among the single measurement variables for this dissertation was .1795 (SD=.1456). 44 Table 2 Correlations among single-measurement variables Beck Anger Arousal Anger at Home Sadness at Home Hassles Uplifts Anger Arousal .5657 p=.ooo **** Anger at Home .3560 p=.005 ** .6024 p=.ooo **** Sadness at Home .4612 P=.000 **** .3026 p=.018 * .3826 E=.002 *** Hassles .2733 .3848 .3491 .0823 p=.035 * p=.002 *** p=.006 ** p=.528 Uplifts -.1564 .0799 .1101 .0543 .3654 p=.233 E=540 B=.398 p=.678 P=.004 *** PMWI Domination -.1261 -.0194 .4027 -.0683 .2233 .0681 - Isolation p=.398 p=.896 E= 005 ** p=.645 E=.127 p=.646 Subscale PMWI Emotional - .0825 .1056 .5729 .2130 .1696 .0893 Verbal Subscale C=.581 p=.475 P=.000 **** C=.146 p=.249 p=.546 Marital Need -.0661 -.0827 -.5007 -.2002 -.0259 .1969 Satisfaction Scale p=.659 B=.576 p=.ooo **** p. 172 B=.861 p=.180 MAQ Bothersome .0669 -.0610 -.1232 -.0498 -.2124 -.3016 fi=.612 p=.640 P=.344 p=.703 B=.10O p=.015* MAQ Debilitating .2587 .2580 .1921 .1580 .0136 -.0001 E=046 * p=.045 * fi=.138 p=.224 fi=.917 p=.999 MAQ Predictable .0852 .1744 .1365 .1060 .0369 -.0319 B=.518 C=.179 B=.294 p=.416 P=.803 p=.807 MAQ Denial -.1292 -.1207 -.1177 -.1159 -.1635 .1265 P=325 p=.354 p=. 366 p=.374 E-.208 P=.331 MAQ Natural .0529 .0747 .1336 .0499 .1762 .0924 p=.688 E=.567 p=.305 p=.703 p=.174 p=.479 Physical Abuse by .2634 .1929 .2707 -.0294 .2758 .0128 a Partner p=.092 B=.215 p.079 p=.852 P=.073 p=.935 Sexual Abuse by a -.0477 .1260 -.1035 -.0662 -.0049 -.1661 Partner E=.762 B=.415 p=.504 B=669 E=.975 p=.281 Emotional Abuse -.0103 .1084 .1394 .1921 .1316 .0868 by a Partner p=.947 p=.478 B=.361 p=.206 p=.389 B=.571 Physical Abuse in .2399 .2016 .2344 .2396 .1292 -.0526 Adulthood p=.072 C=.129 p=.077 p=.070 P=.334 B=695 Sexual Abuse in .1897 .0809 -.0251 .1848 .2418 .0096 Adulthood P=.158 p=.546 p=.851 E=.165 p=.067 B=.943 Emotional Abuse .1188 .3036 .1727 .1148 .2229 .0513 in Adulthood P=.379 p=.021 P=.195 B=.391 p=.093 B=702 Physical Abuse in .2413 .0248 .0840 .1413 -.0515 -.1338 Childhood p=.063 p=.849 p=.520 p=.278 p=.693 B=.304 Sexual Abuse in .2073 .0914 .0440 .2610 -.1115 -.0822 Childhood P=.112 P=.484 P=.737 p=.042 * E=.392 B=.529 Emotional Abuse .3186 .2205 .0089 .2142 .0119 -.1058 in Childhood p=.013 * p=.088 p=.946 p=.097 p=.928 B=.417 45 PMWI Domination-Isolation PMWI Verbal-Emotional Marital Need Satisfaction Inventory M A Q Bothersome M A Q Debilitating M A Q Predictable PMWI Emotional - Verbal Subscale .6594 p=.000 Marital Need Satisfaction Scale -.2787 p=.058 -.6971 p=. 000**** M A Q Bothersome -.0633 D=.669 -.0383 D=.796 -.1853 £=.207 M A Q Debilitating -.0983 p.=.506 .1091 E=. 460 -.1843 E=.210 .1307 P=.315 M A Q Predictable -.0834 p=.573 .0369 p_=. 803 -.0786 p=.596 -.0622 D=.634 .5282 D=.000 **** M A Q Menstruation Should be Ignored -.1075 p=.467 -.1995 p=.174 .2606 p=.074 .0927 E=.478 -.5244 p.=.000 **** -.6364 p=.ooo **** M A Q Natural .1316 p=.373 .1203 fi=.416 .1466 p=.320 -.7092 P=.000 **** -.0521 p=.690 .2909 p=.023 * Physical Abuse by a Partner .3070 p=.045 * .4056 p_=.007 ** .1436 D=.364 -.1214 C=438 .0925 p=.555 .1345 D=.390 Sexual Abuse by a Partner .0503 Q=. 746 .0038 C=. 980 -.0326 £=.836 -.0056 P=. 971 -.1116 p=.471 .2031 D=.186 Emotional Abuse by a Partner .2221 p=.142 .3789 D=.010 * -.2078 D=.176 -.1156 p=.449 .0231 p=.880 .1781 p.= 242 Physical Abuse in Adulthood .1191 p=.436 .3418 p=.022 * -.2670 p=.076 .1072 E=.423 .0888 £=.508 .1017 p=.447 Sexual Abuse in Adulthood .0113 B=.941 -.2266 p=.135 .2556 p=.090 -.1346 rj=314 .1919 p=.149 .2834 p=.031 * Emotional Abuse in Adulthood .1460 p=.338 .2244 p=.138 -.1175 D=.442 -.1430 P=.284 -.0268 p=.842 .0699 fj=.602 Physical Abuse in Childhood -.0578 p=.696 -.0855 p=.563 -.1136 D=.442 .0994 p_=446 .1836 fj=.157 .1051 p_=420 Sexual Abuse in Childhood -.0611 p=.680 .0197 p=.894 .0000 p_= 1.000 -.2918 D=.023 .2057 p=.112 .1708 C=.188 Emotional Abuse in Childhood .0252 rj=. 865 .1906 p=.194 -.2338 p=.110 -.0881 p=.499 .1232 £=.344 .2137 p=.098 Sexual Abuse in Adulthood Emotional Abuse in Adulthood Physical Abuse in Childhood Sexual Abuse in Childhood Emotional Abuse in Adulthood .0687 E=608 Physical Abuse in Childhood .2042 £=.124 .3115 p=.017 * Sexual Abuse in Childhood .3171 B=.015 * .3803 p=.003 *** .3019 B=.018 Emotional Abuse in Childhood .0502 E=.709 .6009 p=.ooo **** .5092 p=.ooo **** .4230 P=.001 *** 47 Type I and Type II Error Type I error occurs when a true null hypothesis is rejected and Type II error occurs when a false null hypothesis is not rejected (Glass & Hopkins, 1996; Westermann & Hager, 1986). Whereas Type I error can be controlled by the experimenter, Type II error is more indirectly influenced through experimental design (Winer, 1971). Although the occurrence of these errors cannot be eliminated, the probabilities of their occurrence can be used to guide in the summary and interpretation of results (Winer, 1971). Traditionally, a stronger emphasis has been placed on controlling Type I than Type II error. With bias against publishing non-significant findings, a Type I error is more likely to be published than is a Type II error, and once published, is less likely to be disconfirmed in subsequent publications (McCarroll, Crays, & Dunlap, 1992). Since the two types of error are inversely related, the strong focus on Type I has meant a lesser focus on Type II. Several researchers have called for greater attention to Type II error and to statistical power (Westermann & Hager, 1986; Winer, 1971). Stevens (1996) has suggested that alpha be raised to .10 or .15 for a smaller sample size and Winer (1971) has hypothesized that .30 and .20 might be more reasonable than the conventional .05 and .01. Winer (1971) has, in particular, argued that in exploratory research, both kinds of errors may be equally important. Assessing the relative "costs" of these two errors is difficult in social science research (Glass & Hopkins, 1996). One primary goal of exploratory research is to indicate variables worth pursuing in future research. Meehl (1990) has noted that Type II error in pilot studies has too frequently led to non-publication and an abandoned line of research. This dissertation was intended to explore the role of a previously unstudied variable in PMS research, and to suggest directions for future research. Although some reviewers have called for tighter control of Type I error in 48 PMS research (Koeske, 1983; Sommer, 1980, 1981; Ussher, 1992), Parlee (1981) has argued that the field of PMS is still "immature" in that theorists differ widely in assumptions, methods, definitions, and variables of importance. She posits that PMS research has not yet arrived at the problem-solving stage which requires a generally shared paradigm. Rather, it is still in a problem-finding, hypothesis-generating period of development, and is thus highly suited to exploratory and correlational approaches. Both Type I and Type II error must be considered in interpreting the findings of this study. Experiment-wise Type I error is a potential concern because of the number of analyses conducted. A Type II error of different magnitude is associated with each possible alternative hypothesis (Winer, 1971). Type II error was a possible concern in this study because of its limited sample size (n=58) and the unknown reliability of the say group variable (Glass & Hopkins, 1996; Stevens, 1996). Null hypothesis statistical testing (NHST) itself has been called into question. More extreme views present the entire tradition of NHST as a mistake (Meehl, 1990), whereas others believe NHST to be valid but extensively misused and misunderstood (Cohen, 1995). The drawbacks of NHST and subsequent controversy over its use first were discussed in 1938 (Berkson, 1938, in Cohen, 1994). Critics of NHST note that although H Q is tested, it is fundamentally unacceptable (Weitzman, 1984). The probability that H Q is true is zero (Cohen, 1994; Meehl, 1990). This criticism has primarily been aimed at null hypotheses of strict equality (Weitzman, 1984), but is nonetheless critical given the conservative use of one-tailed tests. Cohen (1994) has referred to this null hypothesis as the "nil hypothesis". If the nil H Q is always false, then any experimental treatment will lead to statistical rejection if a sufficient sample size (Weitzman, 1984) and sufficiently fine measurement (Baril & Cannon, 1995) are used. Cohen (1994) has stated that 4 9 rejection is inevitable even with small samples. If the nil hypothesis is always false, then the rate of Type I error is more accurately 0 % rather than 5 % and the only error possible is Type II (Cohen, 1 9 9 4 ) . Baril and Cannon ( 1 9 9 5 ) , in contrast, have argued that although theoretically the probability that H Q is true is zero, in reality this is not the case. Although Cohen ( 1 9 9 4 ) has noted that there is no magical formula to replace NHST, researchers have made concrete suggestions to enable a clearer presentation and interpretation of results. In particular, recommendations have focused on providing more information about the precision of estimates and not just relying on point estimates (Winer, 1 9 7 1 ) . Glass and Hopkins ( 1 9 9 6 ) have reminded researchers that statistical significance is not synonymous with practical significance, and have suggested that confidence intervals be used as a method of distinguishing the two. By providing confidence intervals, the magnitude of change is represented, not just the direction of change (Cohen, 1 9 9 4 , 1 9 9 5 ) . The size of the confidence interval depends on measurement error (Frick, 1 9 9 5 ) and is inversely related to sample size (Cohen, 1 9 9 4 ) . Cohen ( 1 9 9 4 ) has proposed that the use of confidence intervals should replace the emphasis on NHST. Parker ( 1 9 9 5 ) has argued for the use of standardized effect sizes rather than confidence intervals. Effect sizes are particularly useful for communicating the magnitude of differences in a common metric (Glass & Hopkins, 1 9 9 6 ; Parker, 1 9 9 5 ) . Effect size indicates the departure of the true parameter value from the parameter value hypothesized in H Q (Glass & Hopkins, 1 9 9 6 ; Westermann & Hager, 1 9 9 6 ) . Cohen ( 1 9 9 4 ) notes that sample effect sizes necessary for significance are typically larger than actual population effect sizes. He warns that a focus on effect sizes may be positively biased. 5 0 Researchers have also emphasized the importance of discussing power in interpreting results (Meehl, 1 9 9 0 ) . If H Q is not rejected, the careful examination of power may reveal the alternative explanation that the experiment did not use a sufficiently sensitive test of the hypothesis (i.e., a Type II error occurred) (Winer, 1 9 7 1 ) . Power is a function of sample size, alpha, and effect size (Glass & Hopkins, 1 9 9 6 ; Stevens, 1 9 9 6 ) . McCarroll, Crays, and Dunlap ( 1 9 9 2 ) have shown, however, that the addition of participants to an already collected sample, post hoc, to increase power and establish difference leads to an elevated Type I error. Type I especially rises when only a few participants are added. They strongly recommend against this procedure. Rather than trying to increase power by increasing sample size, changing hypotheses, or re-designing the experiment post hoc, power and effect size should be reported for both significant and non-significant findings to aid in interpretation. Tukey has claimed that data are rich and complex, and often provide information about unexpected relationships (Borgen & Seling, 1 9 7 8 ) . One recommendation by Cohen ( 1 9 9 4 ) has been to engage in exploratory data analysis. He suggests focusing effort on understanding the sample data instead of generalizing to the population. Generalization relies on replication (Hubbard, 1 9 8 5 ) , rather than on finding a single "best" statistical technique (Borgen & Seling, 1 9 7 8 ; Larabee, 1 9 8 2 ) . Some researchers have proposed the use of a Bonferroni correction to reduce experiment-wise Type I error (Strahan, 1 9 8 2 ) . Although a Bonferroni correction guards against spurious results, power is severely reduced (Stevens, 1 9 9 6 ) . Cohen ( 1 9 9 4 ) has argued that, given that H Q is never theoretically true, Bonferroni procedures are, ironically, adjusting for a non-existent alpha error, while 51 increasing an already high Type II error. A Bonferroni correction was not used for this dissertation. Finally, reviewers have characterized a "conceptual revolution" in statistics that shuns the notion of a single superior statistic (Larabee, 1982). A part of this trend has included a call for more graphic displays of the data to aid interpretation (Borgen & Seling, 1978; Cohen, 1994). Many of the recommendations resulting from the NHST controversy were implemented in this dissertation. Type I and Type II errors have been handled in several ways. For the principal significant and non-significant findings, the means, confidence intervals, effect sizes, rj-levels, and power are reported in tabular form. All of these factors have been considered in assessing the strength, logic, and practical significance of the results. The emphasis has been on an exploratory approach and recommendations regarding variables to be included in future research. Interpretation of the results, therefore, has avoided focusing on singular (especially weaker) results, in favour of the overall picture, to minimize capitalization on chance findings. The emphasis is on patterns of results rather than individual variables. Principal findings are graphed to help demonstrate these patterns. Univariate followups have been conducted only when statistical significance was obtained with MANOVA. The level of Type I error (a) at .05 and the use of two-tailed tests were set a priori for each test. These decisions were adhered to throughout analysis. The Tukey range test was used for multiple comparisons to provide a family-wise error rate and to control Type I error. MANOVA has been used for analysis because it is both a powerful and robust technique. The 15 daily variables were combined into four symptom types for analysis, consistent with Stevens' (1996) suggestion for increasing power and controlling spurious findings. Analyses were not conducted for the individual daily variables. Finally, idiographic 52 analyses have been based on 3-item and 5-item clusters of variables to ensure more reliable units of analysis. Say Group Analyses of Single-measurement Data Do Women's Reports About their Life Circumstances, Emotions, and Attitudes Relate to Their Current Use of the Label "PMS"? A comparison between women who, in the final interview, reported never having PMS (Never-Say, n=23) and those reporting that they currently had PMS (Current-Say, n=35) was conducted for the single-measurement variables. The purpose was to examine the attitudes and life circumstances of women who are using the "PMS" label during their current day-to-day lives. Means, confidence intervals, rj-levels, power, and effect sizes of the parametric significant and non-significant results are presented in Table 3. A 5 (Menstrual Attitude Questionnaire subscales) x 2 (Never-Say, Current-Say) M A N O V A 6 revealed a significant difference on the Menstrual Attitude Questionnaire (MAQ) between the two say groups, F(5,52)=6.54, rj<001. Univariate F-tests indicated that the groups differed on all five subscales of the MAQ. Women who reported never having had PMS were more likely to view menstruation as bothersome (M=28.39), F(1,56)=4.25, rj<05, and were more likely to claim that women's lives should not be affected by menstruation (M=22.74), F(1,56)=11.86, p.<.005, than were women who currently labelled themselves as having PMS (M=24.00 and M= 16.11, respectively) (See Figure 1 ) 7 . Women who 6 In the case of analyses for the two-level independent variable (Never-Say, Current-Say), the MANOVA is equivalent to a Hotelling's T 2 . 7 For figures that are graphed using the box-and-whisker plot (Glass & Hopkins, 1996; Tufte, 1983), all outliers have been plotted. The comparisons for which outliers appeared were re-analyzed with the outliers removed. All results were the same, except for the hassles scale. The results reported in the text are therefore based on the entire sample (that is, including outliers) and the discrepancy in 53 Table 3 Results of parametric analyses of single-measurement variables for differences between Current-Say and Never-Say women Variable Say Group Mean Confidence Interval df Statistic p-level Menstrual Attitude Questionnaire 5,52 F=6.54 p=.ooo **** Bothersome Current Never 24.00 28.39 21.31, 26.69 25.08, 31.71 1,56 E=4.25 0= 044 * Menstruation Should be Ignored Current Never 16.11 22.74 13.68, 18.54 19.75, 25.73 1,56 £=11.86 P=.001 *** Debilitating Current Never 46.86 37.74 41.78, 51.24 32.34, 43.14 1,56 E=6.91 p=.011 * Natural Current Never 27.26 21.04 25.14, 29.38 18.42, 23.66 1,56 E=13.63 P=.001 *** Predictable Current Never 29.37 22.17 27.43, 31.31 19.77, 24.57 1,56 F=21.82 P=.000 **** Anger and Sadness Measures 4,52 E=3.01 p=.026 * Anger at Home Current Never 10.60 8.68 9.30, 11.90 7.08, 10.28 1,55 F=3.39 P=.071 Anger Arousal Current Never 32.43 22.18 28.75, 36.11 17.64, 26.72 1,55 E=12.01 P=.001 *** Beck Depression Inventory Current Never 10.03 6.09 7.83,12.23 3.37, 8.81 1,55 E=4.96 p=.030 * Sadness at Home Current Never 9.80 8.86 8.10, 11.50 7.25, 10.47 1,55 E=-80 p=.375 Hassles and Uplifts 2,55 F=3.04 p=.056 Hassles Current Never .73 .49 .594, .866 .323, .657 1,55 F=4.88 p=.031 * Uplifts Current Never .74 .76 .617, .869 .609, .919 1,55 F=.04 p=.833 Marital Need Satisfaction Scale Current Never 71.94 69.56 69.21, 74.67 66.54, 72.59 45 t=-.39 p=.699 Psychological Maltreatment of Women Inventory 2,44 F=.11 p=.892 Domination-Isolation Current Never 28.55 29.63 25.91, 31.19 26.38, 32.88 1,45 F=.20 p=.654 Emotional-Verbal Current Never 40.97 43.19 35.20, 46.74 36.07, 50.31 1,45 E=.18 E=.672 54 Variable Effect S ize 6 Power 9 Partial Eta Squared (n ) Menstrual Attitude Questionnaire 2 . 5 4 " .99 1 2 Bothersome .55 .58 .57 Menstruation Should be Ignored .93 .95 .17 Debilitating .71 .74 .11 Natural .99 .95 .20 Predictable 1.25 .99 .28 Anger and Sadness Measures .92 .72 Anger at Home .50 .45 .06 Anger Arousal .94 .93 .18 Beck Depression Inventory .61 .61 .08 Sadness at Home .24 .15 .01 Hassles and Uplifts .45 .67 Hassles .60 .66 .08 Uplifts .06 .03 .00 (.0008) Marital Need Satisfaction Scale .39 .31 Psychological Maltreatment of Women Inventory .02 .07 Domination-Isolation .14 .08 .00 (.0042) Emotional-Verbal .13 .08 .00 (.0040) 8 This effect size calculation is presented in standard deviation units, based on the formula A=(M j- M J / s w (Glass & Hopkins, 1996). 9 Post hoc univariate power was calculated using power curves for the non-centrality parameter (Glass & Hopkins, 1996) and power tables for J=2 (Hopkins, Coulter, & Hopkins, 1981) based on a harmonic n (Stevens, 1996). The tables were specifically used to provide a more accurate estimation in cases of low power. 1 0 Cohen (1977) has suggested that the values of partial r | 2 may be interpreted as TI2=.01 small, TI2=.06 medium, r|2=.14 large effect sizes. 1 1 Multivariate effect size calculations for J=2 is based on D 2 , Mahalanobis Distance, and the harmonic n (Stevens, 1996). 1 2 Post hoc multivariate power calculations are derived from D 2 power tables published by Stevens (1996) and on S P S S for Windows output. In the case of a difference, the lowest estimate has been presented here as a conservative estimate of power. 55 S 5 0 " c o r 40-e 30-20-10- I 0. Never-Say Current-Say Menstruation as Bothersome S 5 0 " c o 1 r 40-e 30- , I 20-10-0. Never-Say Current-Say Menstruation Should be Ignored Fig. 1. Menstrual Attitude Questionnaire Subscales on Which Never-Say Women Scored Significantly Higher than Current-Say Women. 1 3 1 3 Data have been graphed using a box-and-whisker plot (Glass & Hopkins, 1996; Tufte, 1983). The box represents the interquartile range of scores, the vertical lines illustrate the range of scores, hollow dots indicate outliers, and the horizontal line is the median score. 56 believed they currently had PMS viewed menstruation as more debilitating (M=46.86), F(1,56)=6.91, £< 05, as more natural (M=27.56), E(1,56)=13.63, rj<.005, and as more predictable (M=29.37), F(1,56)=21.82, p_<001, than did women who had never used the label (M=37.74, 21.04, 22.17, respectively) (See Figure 2). Multivariate and univariate effect sizes, levels of significance, and statistical power for the MAQ were generally strong. In addition, confidence intervals for the two say groups were mutually exclusive for the subscales that measured women's perceptions of menstruation as predictable, natural, and easy to ignore. This suggests that differences between the groups met an especially conservative criterion for these attitudinal scales (Glass & Hopkins, 1996, p. 413, footnote 36). In addition, Figure 1 shows that the interquartile range of scores for the Never-Say women on the debilitating subscale was below the median for the Current-Say women and the effect size for this difference was relatively strong (.71). The difference between the two groups for the bothersome subscale was a slightly weaker finding. Effect size and partial r|2 were moderate, and the level of significance was higher (p_<05). But there was only a slight overlap of the confidence intervals between the two groups, and the statistical power of this subscale was moderate. An increase in power might reveal a stronger effect on this scale in future studies. In this study, the higher endorsement of the natural, predictable, and debilitating subscales by the Current-Say women and the greater characterization of menstruation as something to be ignored by the Never-Say findings for the hassles scale is noted. Outliers occurring on three of the measures were from the same participant. This woman reported particularly high scores for the domination-isolation scale of the PMWI and for the debilitating and natural subscales of the MAQ. In addition, this participant reported emotional and physical abuse by a (current) romantic partner. Although she indicated that she currently had PMS, she did not meet researcher criteria for either PMS or PDD. 57 s c o r 40-30-40-30-20- 20-10- 10-Never-Say Current-Say Menstruation as Predictable Never-Say Current-Say Menstruation as Natural s 90-c 80-0 r 70-e 60-50-40-30-20-10. Never-Say Current-Say Menstruation as Debilitating Fig. 2. Menstrual Attitude Questionnaire Subscales on Which Current-Say Women Scored Significantly Higher than Never-Say Women 58 women who reported currently having PMS were more likely to think of menstruation as an inevitable, biological intrusion that they were able to sense approaching. The women saying that they had never had PMS perceived menstruation as simply bothersome, the effects of which they were able to ignore. Women's reports of anger and sadness in their lives were analyzed using a 4 (anger arousal, anger at home, Beck Depression Inventory, sadness at home) x 2 (Current-Say, Never-Say) MANOVA. Multivariate power and effect sizes for the anger and sadness measures were large, even though p. was not very low, F(4,52)=3.01, £=.026. Follow-up univariate tests revealed significance for anger arousal, F(1,55)=12.01, p_<.005, and the Beck Depression Inventory (BDI), F(1,55)=4.96, p_<.05 (See Figure 3). Women who said that they currently experienced PMS had higher depression scores on the BDI (M=10.03 Current-Say and M=6.09 Never-Say). Both group means on the BDI were below a cutoff score of 15 used by some researchers for screening depression (Bishop & Edgley, 1993; McFarlane & Williams, 1994). The level of significance for the difference between the say groups for the BDI was p_= 030, and the effect size and -n 2 were moderate. Although this finding was not a strong one, this may be due to lower power (1-p=.61). Indeed, the confidence intervals were fairly wide for both groups. A larger sample size might narrow the confidence intervals, increase power, and produce a stronger indication of difference. It is noteworthy, however, that the maximum score for the Current-Say group was substantially higher than that for the Never-Say group and was positively skewed. This reflects the finding that all six women exceeding the clinical cutoff score for depression were in this group. To ensure that depression scores and mood patterns related to menstrual cyclicity were not confounded, the menstrual phase during which each woman completed the BDI was calculated. Beck scores were not significantly related to s 60- 59 c 0 50-r e 1 ~ 40- , 1 ,. 30-20-10-0. Never-Say Current-Say Anger Arousal S c o r e 30-20-10-0. I i Never-Say Current-Say Beck Depression Inventory Fig. 3. Anger and Depression Scales on Which Current-Say Women Scored Significantly Higher than Never-Say Women 60 menstrual phase (rj=.077; means were 3.70 menstrual phase, 6.64 postmenstrual, 8.17 midcycle, 10.20 post-midcycle, and 10.93 premenstrual, with lower scores indicating less depression). Current-Say women had higher anger arousal scores on the MAI (Multidimensional Anger Inventory). The p_-level for the difference between groups for anger arousal suggests that this may be a strong finding (p_=.001). The effect size, power, and partial r|2 for this result were large. Confidence intervals for the two groups were exclusive, indicating that Current-Say women (M=32.43) reported becoming more easily angered, staying angry longer, and experiencing anger more frequently than Never-Say women (M=22.18). The two groups did not significantly differ on the Anger at Home Scale (p_=071; M=10.60 Current-Say, M=8.68 Never-Say). But power was relatively low for the univariate test (.45) of this measure. The effect size was moderate and significance was nearly attained. It is possible that this result represents a Type II error and that significance would be attained with greater statistical power. Figure 4 illustrates more variability among scores for the Never-Say women than for the Current-Say women on this measure, but the median scores for the two groups are roughly the same. Because there was only a 2-point difference between the means and the confidence intervals were 2-3 points wide, practical significance was not assigned to this scale in this dissertation. The Sadness at Home Scale did not approach statistical significance (p_=.375). The group means were close together and the confidence intervals overlapped substantially. Figure 4 shows a similar distribution for both groups. Within this sample, the lack of group difference for this measure seems clear. Power was very low, however, so the lack of significance might not be replicated in a new sample. 61 S c o r e 20-18-16-14-12-10-8-6-4-2-0. Never-Say Current-Say Anger at Home S c o r e 20-18-16-14-12-10-8-6-4-2-0. Never-Say Current-Say Sadness at Home Fig. 4. Anger and Sadness Scales for Which No Significant Differences were Found Between Current-Say and Never-Say Women 62 Further parametric multivariate analyses revealed no significant differences (p_>.05) between women reporting PMS currently and those reporting never having had PMS in how well their romantic partners fulfilled their relationship needs (Marital Satisfaction Scale) (MNS) (p_=.699), and their reports of psychological abuse by their current primary romantic partners (Psychological Maltreatment of Women Inventory) (PMWI) (rj=.892) (See Figure 5). In general for the measures in these analyses, the confidence intervals overlapped substantially and the means were within the confidence intervals for both groups. Therefore, the lack of significant group differences for these measures seem to be unambiguous findings. But caution is necessary in generalizing this result, since power was low for the MNS scale and extremely low for the PMWI. For the Daily Hassles and Uplifts Scale, the multivariate test involving both dimensions demonstrated moderate power, a medium effect size, and nearly attained statistical significance (p_=.056). For Uplifts, which had poor power, the univariate test did not approach significance (£=.83) and the confidence intervals overlapped considerably. But for Hassles, which had moderate power, the mean for Current-Say women (M=-73) was higher than that for Never-Say women (M=49), F(1,56)=4.88, p_<05, as would be hypothesized. Moreover, the confidence intervals for the groups overlapped very little and the effect size for the difference in means was medium. Furthermore, Figure 6 shows that the median score for Hassles for the Never-Say group was below the interquartile range for the Current-Say group. This difference may have practical significance and might reach statistical significance if replicated with a new sample from the same population. There were four outliers for the hassles scale (See Figure 6). Both the multivariate test (Daily Hassles and Uplifts), F(2,51)=4.23, rj< 05, and the univariate test, F(1,52)=7.93, p_< 01, were significant for hassles when the outliers were 63 CO co i CD O CO CO co > CD o o o oo o CO o o CM co o o >> co CO "c cu >> CO CO I > CD CD -£Z 05 CO o co .Q ZJ CO "to .Q l_ CD > i_ to c g o E O J CL CD O _CD co o CO _a CO cz o 00 O o o a >< CO CO c CD i— i ZJ O ca CO I 1— CD > CD 0) CO o CO c g o co co CO •o CD O CN O O o oo o co. o -3- o CM CO O O O i_ CO m O >, 5 ro - c L ^ co r r c | £ CD .2 3 JO O CD CZ Of CD CD •2 5 CO CO or § — o C LL ZJ CD O § CD CO CD O c CD CO O) CD LX CO CD CO cz cz CD E o 5^ CO c S w 2 i Q - ~ CD CO cz > CD O) CD 0- CZ) z o cn o o CO o CM CO o o LO CD 64 S c o r e 1.5-. o o 1.0- 7 .5-0.0. I Never-Say Current-Say Hassles 2.5-o 2.0-S 2.0-c o r 1.5-e 1.0-.5-0.0. Never-Say Current-Say Uplifts Fig. 6. Daily Hassles and Uplifts Scores of Current-Say versus Never-Say Women 65 removed and the data were re-analyzed. These results indicated that Current-Say women (M=-66) reported more frequent and intense hassles than did the Never-Say women (M=-42). Phi correlations were used to assess the relationship between the women's use of the "PMS" label (Never-Say, Current-Say) and their reports of their emotional, physical, and sexual abuse histories in childhood, adulthood, and past romantic relationships. The levels of significance, proportions of women reporting past abuse, and strength of association for these correlations are reported in Table 4. Significant phi correlations indicated that women who currently said that they had PMS were more likely to report having experienced emotional abuse in childhood, phi=.4Q, rj<005, emotional abuse in adulthood, phi=33, £ < 0 5 , and physical abuse by a previous romantic partner, phi=.31, rj<.05. Almost three-quarters (71%) of all the women said that they had experienced some form of emotional abuse during childhood, and 74% of these women were currently using the label "PMS". Slightly fewer than half (47%) reported being emotionally abused as adults. Of these women, 78% said that they currently had PMS. More than a third (38%) of the women reported physical abuse of some form by a previous romantic partner. Of these women, 88% claimed that they currently had PMS. Only 2 said they did not. In contrast, of the women who reported no past physical abuse by a partner, only 58% reported current PMS. Phi correlations did not attain significance (p_>. 05) for ever being sexually abused in adulthood (£=.057; 44% of Current-Say vs. 19% of Never-Say women), being sexually abused by any romantic partner (rj=23), being physically abused in adulthood (£=.135) or sexually or physically abused in childhood (p_=. 159 and £=.106, respectively). Finally, with respect to possible links between relationship variables and use of the label "PMS", recall that more Current-66 Table 4 Results of abuse variables for Current-Say and Never-Say women Variable Say Group Percentage Strength and p-value Reporting Direction of Abuse Association Physical Abuse by a Current 48 Romantic Partner Never 15 phi=.31 p=.042 * Sexual Abuse by a Current 62 Romantic Partner Never 43 phi=.18 E=.235 Emotional Abuse by Current 76 a Romantic Partner Never 67 phi=.10 p=. 516 Physical Abuse in Current 38 Adulthood Never 19 phi=.20 p=.135 Sexual Abuse in Current 44 Adulthood Never 19 phi=.40 p=.057 Emotional Abuse in Current 85 Adulthood Never 48 phi=. 21 p=.003 *** Physical Abuse in Current 26 Childhood Never 9 phi=.21 B=.106 Sexual Abuse in Current 34 Childhood Never 17 phi=.18 B=.159 Emotional Abuse in Current 60 Childhood Never 26 phi=.33 p=.011 * 67 Say (83%) than Never-Say (54%) women reported currently being heterosexually active (phi=.31. rj< 05). Summary of the single-measurement findings. In this study, the woman who said that she currently had PMS was more apt to view menstruation as debilitating, predictable, and natural. She was also less likely to claim that menstruation should not affect women. This woman reported that her anger was more easily aroused and sustained. She was more likely to report having been emotionally abused in childhood. A weaker but statistically significant result indicated that Current-Say women were more likely to report having been emotionally abused in adulthood and physically abused by a romantic partner. Other weaker significant findings suggested that the Current-Say woman perceived menstruation as less bothersome and generally experienced greater (but still subclinical, on average) depression than the woman who said she had never had PMS. A trend was found to suggest that she may experience more hassles than the Never-Say woman. Although this trend did not reach statistical significance at the multivariate level, it was significant in the univariate test. Further examination of the effect size and confidence intervals for this measure suggests that multivariate non-significance might be attributable to Type II error. Women's use of PMS to describe their menstrual cycles was not found to be related to their reports of uplifts, marital satisfaction, psychological maltreatment by their current primary romantic partner, sexual abuse history, physical abuse in childhood or adulthood, emotional abuse by a partner, or sadness or anger experienced at home. The effect size and confidence intervals for anger experienced at home indicate that this variable may be worth investigating in future research. In addition, women's reports of sexual abuse experiences in adulthood correlated moderately (phi=.4Q) to self-designation of PMS, but this result did not 68 quite reach statistical significance (rj=.057). This variable might attain significance in future research. Say Group Normative Analyses of Prospective Data Analyzing symptom types McFarlane and Williams (1994) found no relationship between women's self-designations and researcher-designations of PMS using predominantly mood variables. One purpose of this dissertation was to explore whether these same women's self-designations would be more consistent with their prospective cyclicity for other types of symptoms. It may be that women who believe that they have PMS are basing their decisions on different symptomatology than mood. To investigate this question, the 15 daily variables were conceptually grouped into 4 symptom types: mood (mood pleasantness, mood variability/stability, anger, hurt), relationships (quality of personal relationships, quality of professional relationships, patience with others), concentration and work performance (amount of concentration, quality of paid work performance, quality of domestic work performance), and physical and activity symptoms (balance of body experiences, amount of leisure time, quality of sleep, amount of eating, amount of exercise). To confirm that these categorizations could represent how women might group, and therefore experience, symptoms together, the 15 daily variables were presented to 2 graduate students. The students were asked to group the variables into an unspecified number of similar categories that might represent "types of symptoms". The male psychology graduate student generated 5 categories that were the same as those generated for this dissertation, with the exception that he divided the physical and activity variables into two separate groups. The female sociology graduate student generated 4 categories, in which all but the concentration variable matched the groups outlined above. 69 Are Women's Current Self-desianations of "PMS" Related to Changes Thev Experience on Specific Types of Symptoms? A 4 (symptom type) x 5 (menstrual phase) x 2 (Current-Say, Never-Say groups) MANOVA was conducted to address this question. The confidence intervals, partial T|2, and power calculations for the significant and non-significant results are presented in Tables 5 and 6. A significant symptom type effect, F(2.26, 126.28)=41.89, rj<001, and symptom type by menstrual phase interaction, F(10.25, 574.04)=3.10, £< 005, were found. Further analyses were conducted to reveal the more specific nature of the interaction. These results are shown in Figures 7, 8, 9, and 10. The menstrual phases did not significantly differ for the relationship variables (rj=.165). For concentration and work variables, the menstrual phase (M=5.93) was rated significantly more negatively than all other phases (means are 6.05 midcycle, 6.06 postmenstrual and premenstrual, 6.12 post-midcycle), F(3.24, 184.77)=2.66, £ < 0 5 . For the mood variables, the post-midcycle (M=6.76) and premenstrual (M=6.66) phases were rated most negatively, F(4,54)=3.70, rj<05. The menstrual phase (M=6.81) was rated more negatively than the postmenstrual phase (M=6.92), but more positively than the premenstrual phase. The physical and activity variables were rated most negatively during the menstrual (M=5.71) and post-midcycle (M=5.68) phases, F(4,54)=2.87, rj<05. The premenstrual phase (M=5.75) was rated more negatively than the postmenstrual phase (M=5.84), but more positively than the post-midcycle phase (M=5.68). The premenstrual and menstrual phases did not differ. Physical and activity variables were rated most negatively, followed by concentration and work performance variables, relationships, and mood, across all of the menstrual, F(2.46, 140.37)=41.28, rj<001, (means=5.71, 5.93, 6.51, 6.81, respectively), postmenstrual F(2.49, 142.08)=43.53, p<.001, (means=5.84, 6.06, 70 Table 5 Means and confidence intervals for the prospective variables for the Current-Say and Never-Say women Effect Cel l Mean Confidence Intervals Say Groups Current-Say Never-Say 6.23 6.38 5.37, 7.09 5.32, 7.44 Symptom Type Physical and Activity Concentration and Work Performance Relationships Mood 5.75 6.04 6.56 6.81 5.43, 6.08 5.72, 6.37 6.24, 6.88 6.48, 7.13 Menstrual Phase Menstrual Postmenstrual Midcycle Post-midcycle Premenstrual 6.24 6.37 6.32 6.27 6.25 6.11, 6.37 6.24, 6.50 6.19, 6.45 6.14, 6.40 6.12, 6.38 Say by Symptom Type Range of Means 5.70 to 6.97 Current-Say: M + .42 Never-Say: M ± -51 Menstrual Phase by Symptom Type Range of Means 5.68 to 6.92 M ± .07 Say by Menstrual Phase Range of Means 6.15 to 6.42 Current-Say: M+-17 Never-Say: M ± 2 1 Say by Symptom Type by Menstrual Phase Range of Means 5.63 to 7.09 Current-Say: M + 09 Never-Say: M ± 1 2 71 Table 6 Results of parametric analyses of prospective variables for the Current-Say and Never-Say women Effect df Statistic p-level Partial n 2 1 4 Power Say Group 1,56 E=.90 P=.346 .02 .169 Symptom Type 3,54 E=85.63 E=.000 **** .43 1.00 Menstrual Phase 3.36,188.24 E=2.26 p=.076 .04 .654 Say by Symptom Type 2.26,126.28 E=.39 p=.703 .007 .127 Symptom Type by Menstrual Phase 10.25,574.04 F=3.10 e=. 001 *** .053 .994 Say by Symptom Type by Menstrual Phase 10.25,574.04 F=1.00 p=.444 .018 .591 1 4 For the repeated measures analyses, partial r\2 was the only measure used to determine effect size. Both partial r|2 and power calculations in this table were produced by S P S S for Windows. M 7 0 " e a n 6.5-72 6.0-5.5. M e n s P o s t m e n s M i d P o s t m i d P r e m e n s Relationship Variables M e a n 7.0-6.5-.5.5. M P M P P e 0 i 0 r n s d s e s t t m m m e e i n n d s s Concentration and Work Variables Fig. 7. Results of Prospective Analyses Across Say Groups: Menstrual Phase Comparisons for Relationship Variables and for Concentration and Work Performance Variables M 7 ° -e a n 6.5-6.0-5.5. 73 M P M P P e 0 i 0 r n s d s e s t t m m m e e i n n d s s Mood Variables a n 6.5-6.0-5.5. M P M P P e 0 i 0 r n s d s e s t t m t m m e e i n n d s s Physical and Activity Variables Fig. 8. Results of Prospective Analyses Across Say Groups: Menstrual Phase Comparisons for Mood and for Physical and Activity Variables •M e a n 7.0-6.8-6.6-6.4-6.2-6.0-5.8-5.6-5.4. 74 Mood Cone & Work Rel Phys & Act Menstrual Phase M e a n 7.0-6.8-6.6-6.4-6.2-6.0-5.8-5.6-5.4. Mood Cone & Work Rel Phys & Act Premenstrual Phase Fig. 9. Results of Prospective Analyses Across Say Groups: Symptom Type Comparisons for the Menstrual and Premenstrual Phases o < ofl •= CL O 75 •o8 o c o 0) o >. o •a CD cc o co co co CN CD O CO oo CO CO LO in o < OO cu cc o oq co o < oO • . a) CO CD .C CL "co 3 "co c — m CD c cc £ CO O 0. co co co CN CO O co oo LO <o in CD CO co co o >> o 13 CO O CL CO CD CO CO CD £ C L CL. 0) I S H I c o ^ H o 0 3 CD a) CO (j cn -g co ^ co O _ -CO ZJ < _ co "co CD - £ < O CD 0-> CD o a C L O CO O co * § co ca *= C L 3 s D ? 0 o oo co co CO co CN CO O CO ao co iri iri CD 6.67, 6.92, respectively), midcycle, F(2.68, 152.50)=41.75, £ < 0 0 1 , (means=5.80, 6.05, 6.57, 6.88, respectively), post-midcycle, F(2.25, 128.01)=31.99, £ < 0 0 1 , (means=5.68, 6.12, 6.53, 6.76, respectively), and the premenstrual phases, F(2.31, 131.82)=26.43, £ < 0 0 1 , (means=5.75, 6.06, 6.53, 6.59, respectively). Partial T I 2 for the symptom type by menstrual phase interaction suggested only a medium effect size (.053), even though the level of significance was p_=001. Followup analyses of symptom type differences within menstrual phase demonstrated a consistent ordering of symptom types and strong n, 2 effect sizes across all phases (range=.32 to .43). Interactive differences were most evident for phase differences within symptom types. Effects were generally moderate across the three significant symptom types (n,2=.05 to .06) and smaller for the relationship symptom type (.03), despite variation in levels of significance (Physical and Activity, £=.032; Mood, £=.008; Concentration and Work Performance, £=.045; Relationships, £=.17). Small effects resulted for all other non-significant main and interaction effects of the prospective analyses (See Table 6). Overall, women's self-designations of PMS were not related to any systematic pattern of symptom types. In other words, women who said that they currently experienced PMS did not show evidence that they cycled differently than women who said that they had never had PMS for any of the mood, physical and activity, relationship, or concentration and work performance symptom classifications. These findings suggest that women may not be basing their different decisions to use "PMS" on the different types of symptomatology that they experience. It also indicates that the discrepancies observed by McFarlane and Williams (1994) between self-designations and researcher-designations probably did not occur because women were focussing on a type of symptom other than mood. 77 Examining How Women Who Sav Thev Have PMS Are Using the Label The analyses just described focussed on women who said they currently had PMS and those who said they had never had it. The 35 women who said that they currently had PMS included 16 who said that they had always had it and 19 who said that they had it now but had not always had it. There were, in addition, 3 women who said that they had had PMS in the past but did not currently have it. in the hope of shedding light on factors related to women's use of the "PMS" label, the analyses were re-run for three groups, based on the consistency of their use of the label "PMS" over their menstrual years: Always-Say (n=16, M=34.94 years), Sometimes-Say (n=22, M=35.18 years), and Never-Say (n=23, M=33.43 years). Do women's reports about their life circumstances, emotions, and attitudes relate to the consistency with which they use "PMS"? The 3 Say groups of women were compared first for the single-measurement variables. The significant and non-significant results of the parametric analyses of these variables are shown in Tables 7 and 8. A 5 (Menstrual Attitude Questionnaire subscales) x 3 (Always-Say, Sometimes-Say, Never-Say) MANOVA was significant, F(10,100)=3.16, p_< 005. Univariate F-tests were significant for all five subscales. Results supported the 2-Say group findings that women who described themselves as having PMS were more likely to view menstruation as natural (M=27.19 Always-Say, M=26.00 Sometimes-Say), F(2,58)=5.16, p_<01, as predictable (M=28.88 Always-Say, M=29.64 Sometimes-Say), F(2,58)=11.34, p_<001, and more likely to minimize the effects of menstruation on their own and others' lives (M=16.31 Always-Say, M=15.73 Sometimes-Say), E(2,58)=6.62, p_< 005, than were women who believed that they had never had PMS (M=21.04, 22.17, 22.74 respectively). Interestingly, the 3-Say group analysis revealed that it was the women who believed that they had always had PMS (M=21.88) who were less likely than the Never-Say women, 78 Table 7 Means and confidence intervals for the single-measurement variables for Never-Say, Sometimes-Say, and Always-Say women Variable Say Group Mean Confidence Interval Menstrual Attitude Questionnaire Bothersome Never 28.39 25.10, 31.68 Sometimes 26.77 23.41, 30.14 Always 21.88 17.93, 25.83 Menstruation Should Never 22.74 19.80, 25.69 be Ignored Sometimes 15.73 12.72, 18.74 Always 16.31 12.78, 19.84 Debilitating Never 37.74 32.36, 43.12 Sometimes 48.50 43.00, 54.00 Always 45.81 39.36, 52.26 Natural Never 21.04 18.32, 23.76 Sometimes 26.00 23.21, 28.79 Always 27.19 23.92, 30.46 Predictable Never 22.17 19.79, 24.55 Sometimes 29.64 27.21, 32.07 Always 28.88 26.03, 31.73 Anger and Sadness Measures Anger at Home Never 8.68 7.02, 10.34 Sometimes 10.27 8.58, 11.96 Always 10.56 8.57, 12.55 Anger Arousal Never 22.18 17.59, 26.77 Sometimes 32.14 27.45, 36.83 Always 31.25 25.75, 36.75 Beck Depression Never 6.09 3.41, 8.77 Inventory Sometimes 11.09 8.35, 13.83 Always 7.19 3.98, 10.40 Sadness at Home Never 8.86 7.32, 10.40 Sometimes 10.77 9.19, 12.35 Always 8.44 6.59, 10.29 Hassles and Uplifts Hassles Never .492 .329, .655 Sometimes .776 .609, .943 Always .688 .493, .883 Uplifts Never .764 .607,.921 Sometimes .732 .571,.893 Always .853 .665,1.041 Marital Need Never 69.56 61.23, 77.89 Satisfaction Scale Sometimes 72.53 64.01, 81.05 Always 70.80 60.81, 80.79 Psychological Maltreatment of Women Inventory Domination-Isolation Never 29.63 26.31, 32.95 Sometimes 28.72 25.32, 32.12 Always 29.14 25.17, 33.11 Emotional-Verbal Never Sometimes Always 43.19 42.39 39.93 36.06, 50.32 35.10, 49.68 31.38, 48.48 80 CO _ Q CD CO CO CD E o CD _ CO CD CO > ^ <-< CD E < CD 5§ $> CD E CO 4& E i o E Q) CO CO -> ^ 5 ^ CD TO ™ CO CD tL O S> i= CD <U Z ro o o. E >4- ro ° CO P" "co co CD o D_ £ N UJ CO CD > CD O .52 co CO c o co co Q . E o O jD .O CO l_ CO > OO CD O o 61 co co II, 5 co TO g 5 as .S T} co 3 to C ^ CD CD £ 3 § •< O CD CO T - CM CO CM CO CM O al LO co co illj CO CD E CO -^3 >. CD i E < co > > CD CD > > CD CD CD E o w CD . c o CQ O r~ CO CO cn CD o co o o ol CM CD CO CO CO CD •B "CD p E C O . > > co i— •> 5 CO 6 1 CD . Q CO co -3 c o o CD £ c , S co ^ co co T -CO 00 CM o CM q 61 o CM CO CD E CO ^ CD i E .5 o < CO > > 1_ I CD CD > > CD CD CO CD CD Q LO "3" CO CO O) N CD O O Ol LO CO CD E CO U 3 > N CD i E .£ o < CO > > CD CD > > CD CD CO CN 00 CO CO o o o CO CO CD II £ 3 CD p E 5 ° . > > co, 9 & z < CO CO o 61 00 •o £ w fD CO W i->- ® 3 ( C C S C CO 03 <c CO § co o CD co 3 .Q CD | g £ : 3 c « 2 s 81 o CM r>- o o CD co T - oo CO CD o LO CD N 00 r-T - r-- co o LO CN CO LO CD CD O CD LO O CM CN LO I— CN CN O co CO 00 CM CN O CO 1^-o o LO CD o CD LO CD O T - O 00 O O O CD O o LO o CD LO -*t T ~ O T -I CN Ik 00 o o ol CO CO o 61 Q i "3-CD o ll'l al CD LO o CD CD CD CO CD ol co CD I CO Q l CN CO ill CM co LO CM co co ib CD CN CM CO o CN LL.I co o co ib CD Lij CO CD O " I LL.I LO LO o ib 00 LO CN LO_ CN 1^ -LO CN LO CN CD 00 LO CN 00 LO CN LO co CD LO CN LO CN" CO :z; >. 0 CD c iJ I < CO co CD E co >. 0 i E > o < CO co 0 E CO -J3 CD i E > o < CO > > 1_ I CD CD > > CD CD > > CD CD > > CD CD > > CD CD > > CD CD co co CD | . i •J=. CD P E . > > CO CO CD E CO ^3 >> CD i E > o < CO > > 1 I CD CD > > 0 0 CO 0 E CO ^3 >* 0 i E > o < co . > > > CO ' - ' - > • 0 0 co > > > 0 0 > CO CO 0 §.i ^ -I—' 0 p E 5 ° c o ^ . > > CO CO 0 E CO ^ 3 > N 0 i E < CO 0 0 > > 0 0 co 0 E -*—* 0 E o co CD < CO CO 0 II •JZ 0 p E E o . > > CO 0 J •z. < 0 E ro 0 CD 1^  0 D3 c < g CO CO 0 1 CL 0 Q o 0 CO o cz 0 > CO I T3 C CO CO 0 "co CO 5 0 .o CO C L ZD co 5 co co -fe: O CO •c cp £ ^ 5 F II CD E -S o o CD c o o E ILU co n i_ 0 > 82 (M=28.39), F(2,58)=3.35, p<.05, to view menstruation as a bother. Women who believed that they had experienced PMS intermittently over their menstrual history (Sometimes-Say) (M=48.50) viewed menstruation as more debilitating, F(2,58)=4.20, p_<05, than did women who believed that they had never had PMS (M=37.74) (See Figure 11). Statistical power was generally good for the multivariate and univariate tests for the MAQ. The strongest findings in these analyses occurred for the subscales characterizing menstruation as natural, predictable, and something to ignore, for which both groups saying they had PMS differed from the Never-Say women. The rj-levels were very low (p_=009, p_= 000, p_=.003, respectively). Effect sizes were large for the groups that significantly differed and negligible for those that did not. Results were weaker for the other two subscales that demonstrated differences related to how women used "PMS". The rj-levels were not particularly low for either the bothersome subscale (p_=.042) or the debilitating subscale (p.=.020). Effect sizes were large for the significant pairwise comparisons for both of these subscales. A medium effect size but lack of statistical significance occurred between the Always-Say and Never-Say groups for the debilitating subscale. This pattern is similar to the subscales measuring perceptions of menstruation as predictable, natural, and something to ignore. In contrast, the bothersome subscale demonstrated a medium effect size between the Always-Say and Sometimes-Say groups. The difference did not attain statistical significance and power was relatively low for this subscale. It may be that women who say that they have PMS vary in their perceptions of menstruation as bothersome depending on how consistently they view their experience, but that this difference was undetectable in this study. This might help explain why the difference between the Current-Say and Never-Say groups was not as strong for the bothersome subscale as for the other MAQ subscales. 83 S c o r e 90-80-70-60-50-40-30-20-10. Never-Say Sometimes-Say Always-Say Menstruation as Debilitating S c o r e 50-40-30-20-10-0. Never-Say Sometimes-Say Always-Say Menstruation as Bothersome Fig. 11. Subscales of the Menstrual Attitude Questionnaire for Which 3 Say Groups Findings were Different from 2 Say Group Findings 84 A significant 4 (anger-arousal, Beck Depression Inventory, anger at home, sadness at home) x 3 (Always-Say, Sometimes-Say, Never-Say) MANOVA, E(8,110)=2.20, rj<05, revealed that, of the women who labelled themselves as having PMS, only the Sometimes-Say women (M=11 09) had higher scores on the Beck Depression Inventory than did the Never-Say women (M=6.09), F(2,57)=3.62, P_<05 (See Figure 12). Although the level of statistical significance was not particularly small for the.multivariate test, the effect size for this pairwise comparison was fairly strong (.78). Neither of these groups differed significantly from the Always-Say group (M=7.19). The effect size between the Never-Say and Always-Say groups was small and confidence intervals for these groups overlapped considerably. The effect size between the Always-Say and Sometimes-Say women for the BDI, however, was moderate (.61). The confidence intervals for these groups were wide, but did not overlap much. Statistical power for the univariate test of the BDI was .65. It is plausible that a larger sample and greater statistical power would narrow the confidence intervals and demonstrate a significant difference between the two groups of women who say that they have PMS. Results for anger arousal supported the findings of the 2-Say group analyses. Women who said that they had PMS (M=31.25 Always-Say, M=32.14 Sometimes-Say) reported that they were more easily angered and felt angry longer than did women who said that they had never had PMS (M=22.18), E(2,57)=5.32, p_<01. Effect size and partial n,2 were large and power was good for the univariate test. The effect sizes for the significant pairwise comparisons were large and the confidence interval for the Never-Say women was lower than that for both of the other two groups. Effect size between the Always-Say and Sometimes-Say groups was very small. Moreover, the means for these groups were nearly the same. Overall, the findings for anger arousal were both strong and clear. 85 S 30-C o r e 20-10-0. I I • — J — Never-Say Sometimes-Say Always-Say Beck Depression Inventory Fig. 12. Depression Scores for the Never-Say, Always-Say, and Sometimes-Say Women 86 In contrast to the 2-Say group result, the anger at home measure did not approach significance for the 3-Say group analyses (p.-274). Power was low (.27) for the univariate test and the overall -n 2 and pairwise effect sizes were relatively low. Confidence intervals for the 3 groups overlapped considerably. The sadness at home scale did not yield a significant group difference (p.= 110), but power was low (.45) for this test. The mean for the Sometimes-Say women was slightly higher than for the other two groups, characterized by moderate effect sizes and -n 2. But the confidence intervals overlapped considerably. The 3-Say group findings for anger at home and sadness at home were somewhat inconsistent with the results of the 2-Say group analyses, in which anger at home demonstrated a greater effect and was close to significance. This inconsistency suggests that the findings related to these two scales may be unreliable. As with the 2-Say group comparison, no significant differences were found among the 3-Say groups on the Daily Hassles and Uplifts Scale (p_=.094), the Psychological Maltreatment of Women Inventory (p_=.892), or the Marital Need Satisfaction Scale (p_=.911). Whereas for the 2-Say group findings, the difference on the hassles scale was significant, for the 3-Say analysis it only approached significance (rj=.056) on the univariate test, with the greatest effect size (.72) occurring between the Never-Say (M=-492) and Sometimes-Say (M=-776) women, and only a medium effect size occurring between the Never-Say (M=-492) and Always-Say (M=-688) groups. The Sometimes-Say and Never-Say confidence intervals overlapped very little, but the confidence intervals of the Never-Say and Always-Say groups were not as clearly separated. The means and confidence intervals for the three groups for the uplifts scale were very close and effect sizes and r| 2 were generally low, as was true for the two-group comparison. Power was 87 limited for the multivariate (.60) and univariate tests (1-p=.56 hassles, 1 -(3=. 13 uplifts). It is possible that a Type II error occurred in detecting differences for Hassles between the Never-Say group and the two groups of women saying that they had experienced PMS. This conclusion is consistent with the results of the 2-Say group analysis. Effect sizes and power were very low for both the Psychological Maltreatment of Women Inventory and Marital Need Satisfaction scales. For both measures, group means were close and confidence intervals showed a great deal of overlap. With such low power, the finding that women's use of the label "PMS" was not related to these two scales for this sample cannot be confidently generalized. Chi-square tests were used to evaluate the relationship between the 3-Say groups and the women's emotional, physical, and sexual abuse histories reported for adulthood, childhood, and past romantic relationships.1 6 The significant and non-significant results of these analyses are in Table 9. Women's reports of emotional abuse in adulthood, x2(2)=8.54,p_<05, were related to whether they used the "PMS" label or not rather than how they used it (48% Never-Say, 81% Always-Say, 86% Sometimes-Say reported abuse). The 2-Say group analysis had revealed a link between both the history of physical abuse by a partner and women's tendency to say they had PMS that was not significant in the 3-group analysis (p_=.134). In addition, the link between women's reports of emotional abuse in childhood and self-designations of PMS was statistically significant for the 2-Say 1 6 Glass and Hopkins (1996) have cited evidence by Roscoe and Byars (1971), by Conover (1974), and by Camilli and Hopkins (1978, 1979) that chi-square is an accurate and effective technique with an average expected frequency as low as 2 (p. 335, footnote 20). Each of the chi-squares used here was based on 58 participants distributed over 6 possible cells. For all significant chi-squares, traditional assumptions of a minimum expected frequency greater than 1 and fewer than 20% of cells with an expected frequency below 5 were met. 88 Table 9 Results of non-parametric analyses of abuse variables for Never-Say, Sometimes-Say, and Always-Say women Variable S a y Group Percentage Reporting Abuse Statistic p-value Physical Abuse by a Romantic Partner Never Sometimes Always 15 50 42 X (2)=4.01 p_=. 134 Sexual Abuse by a Romantic Partner Never Sometimes Always 43 56 75 X (2)=2.74 p=.254 Emotional Abuse by a Romantic Partner Never Sometimes Always 67 78 75 X (2)=0.54 p=.763 Physical Abuse in Adulthood Never Sometimes Always 19 43 31 X (2)=2.78 p_=249 Sexual Abuse in Adulthood Never Sometimes Always 19 48 38 X (2)=3.88 p=.143 Emotional Abuse in Adulthood Never Sometimes Always 48 86 81 X (2)=8.54 p=.014 * Physical Abuse in Childhood Never Sometimes Always 9 27 19 X2 (2)=2.63 p=.268 Sexual Abuse in Childhood Never Sometimes Always 17 36 25 X(2)=2.11 p=.348 Emotional Abuse in Childhood Never Sometimes Always 26 59 56 X (2)=5.87 p=053 89 group analysis, but marginally non-significant when comparing three groups (p_=.053). Overall, women's reports of history of abuse showed no meaningful relationship to the consistency with which they characterized themselves as having PMS. Does the consistency with which women use "PMS" relate to their prospective cyclicity? A 4 (symptom type) x 5 (menstrual phase) x 3 (Always-Say, Sometimes-Say, Never-Say groups) MANOVA revealed no significant main (p_=572) or interaction (p. ranged from .466 to .719) effects for the 3-Say groups on cyclicity of symptom types. Partial -n 2 effect sizes were low across all analyses. Power was moderate for the 3-way interaction, but low for the 2-way interactions and main effect analysis (See Tables 10 and 11). Consistent with the Current-Say and Never-Say comparison, the 3-Say group analyses showed no classic PMS pattern of cyclicity for women who believed they had PMS, regardless of how they were using the label. Normative analyses of the say groups did not reveal differences in prospective cyclical patterns in a way that would explain women's use of "PMS", so idiographic analyses were conducted to examine the cyclical patterns of each individual woman for evidence of a PMS pattern and to explore more closely the nature of each woman's menstrual pattern. Say Group Idiographic Analyses of Prospective Data Were Women's Self-designations of PMS Related to Researcher-designations of PMS? The prospective data were examined to determine whether the discrepancy between women's and researchers' desigations of PMS found by McFarlane and Williams (1994) might have occurred because women saying that they had PMS 90 Table 10 Means and confidence intervals for the prospective variables for the Never-Say, Sometimes-Say, and Always-Say women Effect Cell Mean Confidence Intervals Say Groups Never-Say Sometimes-Say Always-Say 6.38 6.30 6.18 5.26, 7.50 5.22, 7.38 4.57, 7.79 Say by Symptom Type Range of Means 5.70 to 6.97 Never-Say M+-51 Sometimes-Say M±-52 Always-Say M_±-61 Say by Menstrual Phase Range of Means 6.12 to 6.46 Never-Say M + -21 Sometimes-Say M + -21 Always-Say M ± .41 Say by Symptom Type by Menstrual Phase Range of Means 5.60 to 7.09 Never-Say M+ 1 2 Sometimes-Say M+-12 Always-Say M+-14 91 Table 11 Results of parametric analyses of prospective variables for the Never-Say, Sometimes-Say, and Always-Say women Effect df Statistic p-level Partial T I 2 Power Say Group 2, 58 F=.56 p=.572 .019 .142 Say by Symptom Type 4.68, 135.66 F=.92 p=466 .031 .357 Say by Menstrual Phase 7.17, 207.83 E=.67 p=.703 .023 .307 Say by Symptom Type by Menstrual Phase 21.05, 610.44 F=.80 P=.719 .027 .698 92 were placing importance on a different type of symptom than the mood symptoms they reported. The normative results described above indicated that women who said that they had experienced PMS did not differ in cyclicity from those who said that they did not for four different types of symptoms: mood, concentration and work performance, physical and activity, and relationship. It seemed possible, however, that the 15 variables might combine in very individual ways to form menstrual patterns that are consistent with women's self-labelling. In particular, women might be basing their self-designations on a mixture of premenstrual symptoms, rather than on a particular symptom type. To investigate this possibility, we examined each woman's menstrual cycle for evidence of a PMS pattern using all 15 variables. The women were assigned to researcher-designated "PMS" or "nonPMS" and "PDD" or "nonPDD" designations (See Tables 12 and 13). A comparison of PMS researcher-designations and PMS self-designations. Nine women (16%) met the criteria for PMS classification based on the McFarlane. and Williams (1994) method. These women were approximately equally distributed into those who said they were experiencing PMS currently (n=5) and those who said that they had never experienced PMS (n=4) (See Table 14). This result would suggest little difference in cyclical patterns between women who did and did not use the label "PMS". When the Current-Say group was examined more closely in relation to the consistency with which "PMS" was applied, however, a more distinct pattern appeared. Only 1 (6%) of the Always-Say women, but 4 (18%) of the Sometimes-Say women met PMS criteria. None of the women who reported having PMS in the past showed evidence of PMS cyclicity currently, based on the researcher designation. Two women who had a premenstrual M were not classified as having PMS because they had another M cluster during another phase. In both cases, the non-93 r o i _ -4—' co c CD E CD c o 0 o >< o T3 0) CO CO CD T3 c co CO c "co c 0> CO CD " O Hi CD CD (O CO - C CO 3 2 co D_ E CO c CD E co CD CO CD - Q CO "co o E CD CO CO CO 2 2 CD n CO CO O CL j D O > , o • o CO I "oo c CD E "co o 0-c o CO c CD co c D ) CO CD TD M » CD CO CO " c CD E o T3 c co CD CO I- c CD O ^— -4—• o c o CD .S" CO CO CD CD C O C O C O C O C O C O C O C O C O C O C O C O C O C O C O S S S S S 2 S S 2 S 2 S C O S 2 S Q _ Q _ Q _ C L Q - C L C L O _ Q - C L Q _ C L ^ C L Q _ Q -C C C C C C C Z C C C C C Q _ o o o o o o o o o o o o c c c c c c c c c c c c o o o C C C C M C O O C ) C S S T - C M C O C n T - C D C O S t - C N o c N i n ( D N s c o c o a ) c o n ^ c o r - - c o o c o c o c o c o c o r o r o r o m c O T r ' t ^ - ' ^ ^ - u o C O C O C O C O C O C O C O 2 0 - 0 . 0 . 0 - ^ 0 . 0 . 0 . L J - /"> ^ s *-L-o o o o c c c c c c c o o o c c c c o - ^ - o i c D C D r - T - c o c N O T - C N - ^ - O O C D O O I -c 5 co O § CO ^ 3 c 55 E • -o CO O CD 5 D > * co C O I CO >> co < co c C O fD CO CD E "S E o CO 3 o 94 CO CO CO CO CO LO to CO C 0 S 2 2 2 S C O 2 2 2 2 0 . 0 . 0 . 1 1 0 . 2 0 . 0 . 0 . 1-1- /—\ / - \ / - \ /—\ 1-1-o o o o o c z c z c z cz: c o o o c z c c z co co co cocococococococococococo co co co co co co co Q . Q . Q . 2 Q - O . O . Q . Q . Q . Q . Q . O . Q . Q . Q . 2 Q - 2 Q - Q - Q . 2 Q - Q - D -o o o c c c z o o o o o o o o o o o o t z c z c c c c c z c z c z c z c z c o c o o o c z cz: c z c z c z c o o o c c c i - C N ^ - ^ L O C D t O C O C O O ) c z O " c o c o CL c o c D o ^ o a i L o c o f f i C N N c n o i N ^ o i c D O N ^ N T - L n n o o r o i - N T - O T - r - C O M n i n c O O ) 0 0 ( N C N n ^ L O ( D ( D N N C O C I 1 0 c o ^ L o m c o c r i c o m n c o c o m ^ ^ ^ ^ ^ r t ^ ' ^ ^ ' t T f ^ f T t i n CO CO CD > CD 9 5 co ZJ -4—' CO cz CD E CD o zs . a as o o •o aj co CO CD •o cz co a" Q CL co c o '-*—> CO cz O ) CO CD • D iL CD .£Z o co CD CO CD i c o CO cz o "co c .gj CO CD • a I a5 C O C O "c CD E o CD co CO co ZJ •*-» co CZ CD E >. JZ> CO 1 CD "co E CD "T LO _CD JZl co i CD CO H= E T3 c co o CO o Q . _CD O >* o " D co ZJ l_ "co cz CD E co o CL co ZJ —* CO cz: CD CO CD o O CO cz CO CO CO CD CD CH a a a CL cz o cz. Q a a a a a a. a. cx cz c cz o o o c cz c a a a a o D CL CL Q . cz c cz o o o c c cz a a a Q a. Q . c cz o o c c Q Q CL cz: o cz Q Q Q Q Q Q Q 0 , Q - 0 CL CL CL D_ Q . cz cz cz cz cz o o o o o c c cz cz cz a a a a CL a_ cz cz O O CZ CZ ( N C O O C J ) ( N N T - ( N n O ) ' - t O O O S T - C M O C N I L O I D N N C O C O C O O O O ' ^ C D S O O O c r J c o c o c o c o c o c o c o c o c O ' ^ - ^ - r t ^ - ' r i - u n 0 0 - v t / t N C O C D r - ^ i - C O C N 0 1 - N ^ C O O ) O O T -n c o n c o m m T f T t ' * c O CO "cz CD CO cz. o "co c D ) CO CD Q co CO I co >> co 5 < co 2 co CD E -*—« CD ZJ O o CO 96 Q Q Q Q D_ D_ c c o o c c Q Q Q Q D_ Q . c c o o c c Q Q Q Q Q Q CL CL CL o o o c c c Q Q Q . c o c a a a a Q a Q a a. a. c c o o c c a a a a Q a c c c o o o c c c D Q D Q Q Q D D CL D_ CL D_ c: cz cz c o o o o c: c: c: c n t o o ' t o f f i i f i c o o i c N s r o o c N T f m c o o s T - N T - L n w c o a i T - N T - o i - T - t o n n m c o r o o o t M C M n ^ i n t D i O N M a f f i o c o ^ u n c o c o c o c o c o c o c o c o c o ^ ^ - r r ^ - ^ T t - ^ ^ T ^ co CL co 00 1^ CD > 0 o-Table 14 Number of Current-Say and Never-Say women meeting researcher-designated PMS criteria Never-Say Current-Say nonPMS 19 30 PMS 4 5 Number of Always-Say, Sometimes-Say, and Never-Say women meeting researcher-designated PMS criteria Never-Say Sometimes-Say Always-Say nonPMS 19 17 15 PMS 4 4 1 98 premenstrual M cluster occurred menstrually. Both women said that they currently had PMS. One of these women said that she had always had PMS and one said that she was experiencing PMS intermittently. If this restriction was lifted, 19% of the total sample and 20% of the Current-Say group would meet researcher criteria. Overall, there was little correspondence between women's self-designations and researchers' designations of PMS. About 85% of women saying that they currently had PMS failed to meet the researcher criteria. Fifteen of the sixteen women who said that they had always had PMS showed no prospective PMS pattern. Almost four-fifiths (79%) of those who said that they intermittently experienced PMS and said that they had it currently were classified as nonPMS. This was very similar to the proportion of Never-Say women (83%) who were designated as nonPMS. PDD and nonPDD classifications. Five women (9%) met the stipulations for PDD classification based on the DSM-IV criteria (See Table 15). Two of these women believed that they had always had it (Always-Say), two said that they had never had it (Never-Say), and one felt that she had it intermittently and currently (Sometime-Say and Current-Say). With such small numbers, it is impossible to make comparisons among the Say groups. It is noteworthy, however, that the PDD women fell in all three Say groups. Ten women in this study (16%) were designated as either PMS or PDD, or as both. Overall, these designations did not reliably match women's own decisions about whether they had PMS. The DSM-IV distinguishes between PMS and PDD by defining PDD as affective and PMS as more prevalent. All of the women who met researcher criteria for both PMS and PDD had at least one negative mood symptom, consistent with PDD requirements. The incidence rate of PDD in this study approximated that of PDD in the DSM-IV. The incidence rate of PMS in this 99 Table 15 Number of Current-Say and Never-Say women meeting researcher-designated PDD criteria Never-Say Current-Say nonPDD 21 32 PDD 2 3 Number of Always-Say, Sometimes-Say, and Never-Say women meeting researcher-designated PDD criteria Never-Say Sometimes-Say Always-Say nonPDD 21 20 14 PDD 2 1 2 100 study was only slightly higher, but this is not to argue that the PMS-PDD distinction in this study was identical to that of the DSM-IV. It is not the distinction in wording that is of interest, but rather whether stricter criteria better reflect women's own understanding of their premenstrual difficulties. Neither research-based classification reflected women's own notions of whether they were experiencing a Premenstrual Syndrome. PMS, PDD, and Say group classifications for women taking oral contraception and for women with clinical BDI scores. Nine women reported taking oral contraception at the beginning of this study. Eight did not meet PMS or PDD researcher criteria, but 5 believed that they currently had PMS. One woman was taking oral contraception, said that she had never experienced PMS, and met PMS but not PDD researcher criteria. Six women were included whose BDI scores exceeded 15 (the range was 19 to 28, M=23.5). All women said that they currently had PMS. None met researcher criteria for PMS and one met researcher criteria for PDD. One had no T or M clusters for any menstrual phase, one had M clusters premenstrually and menstrually, and two had M clusters menstrually. It is noteworthy that the only woman in the study who had M clusters for 3 menstrual phases (menstrual, post-midcycle, and premenstrual) exceeded the cutoff score on the BDI (her score=25). The relationship among prospective menstrual patterns, self-designations of PMS, and depression is addressed in the Discussion section of this dissertation. Were women's self-designations related to the amount of cyclicity they experienced? According to the McFarlane and Williams (1994) 3-item criteria, most women experienced either one (41%) or two (29%) significant T or M clusters across menstrual cycle phases (see Tables 16 and 17). Only one woman had three Table 16 Distribution of terrific and miserable clusters for Current-Say and Never-Say self-designations Number of Percentage of women with T or M clusters 3-item clusters Current-Say Never-Say Total (n=35) (n=23) (n=58) 0 1 2 3 4 31 37 29 3 0 21 48 30 0 0 28 41 30 2 0 Number of 5-item clusters 0 1 2 3 4 71 23 6 0 0 70 30 0 0 0 72 25 3 0 0 Table 17 Distribution of terrific and miserable clusters for Always-Say, Sometimes-Say, and Never-Say self-designations Number of Percentage of women with T or M clusters 3-item clusters Always-Say Sometimes-Say Never-Say Total (n=16) (n=22) (n=23) (n=61) 0 44 23 21 28 1. 38 36 48 41 2 19 36 30 30 3 0 5 0 2 4 0 0 0 0 Number of 5-item clusters 0 75 73 70 72 1 13 27 30 25 2 13 0 0 3 3 0 0 0 0 4 0 0 0 0 103 clusters - one menstrual T and a M in each of the postmenstrual and post-midcycle phases. This participant said she had PMS currently and sometimes. Differences between the proportions of Current-Say and Never-Say women who had 0, 1, 2, 3, or 4 M or T clusters in at least one menstrual phase were analyzed using a z-test of independent proportions (Ferguson, 1959; McClave & Benson, 1982). No significant differences were found between these two groups (p>.05). Based on the DSM-IV 5-item criteria, a quarter of the women still had at least one cluster (25%), but the number of women with two clusters dropped substantially (3%). By the McFarlane and Williams (1994) criteria, cyclicity is normative, characterizing approximately 80% of each of the Sometimes-Say and Never-Say women, but only 56% of Always-Say women. For two-thirds of the Always-Say women, the cyclicity that did occur fell in only one menstrual phase. For both the Sometimes-Say and Never-Say groups, half the women showed a cluster in only one phase, but half showed a cluster occurring in two phases. One Sometimes-Say woman had T or M clusters in three different phases. Overall, the number of clusters was related more to the way in which women used the label "PMS" than to its mere use or non-use. By DSM-IV criteria, cyclicity was not normative. Close to three quarters of the sample did not have a M or T 5-item cluster. This proportion was distributed across all say groups. The vast majority of the women (88%) who had a 5-item cluster had only one, in a single menstrual phase. Two women had two 5-item clusters, both of whom said that they had always had PMS. Both the 3-item and 5-item criteria were based on the use of prediction analysis to measure "abnormal" cyclicity for the individual relative to her own mean. At the group level, the 3-item criterion suggests that most women experience some 104 cyclicity that is abnormal for them. The 5-item DSM-IV criterion requires that an individual experience "abnormal" cyclicity for a greater number of items. Based on this criterion, a minority of women (28%) experienced cyclicity that was abnormal for them. Therefore, what constitutes "normal" and "abnormal" cyclicity for an individual relative to the population of women may differ quite substantially depending on the criteria that the clinician selects. It would appear, based on this sample, that the use of a 5-item prediction analysis procedure would be a better measure of "abnormal" cyclicity relative to the population. But neither the 3-item nor the 5-item criteria matched women's self-designations and neither provided evidence for a classic PMS pattern either within or across say groups. Were women's self-designations related to the distribution of cyclical clusters over their menstrual cycles? McFarlane and Williams (1994) asked why the premenstrual phase was emphasized so strongly diagnostically when nonpremenstrual cyclicity was more prevalent in their results. In this study, this question was addressed by examining the occurrence of M and T clusters in each of the menstrual phases. Of particular interest was whether the presence of unusually positive or negative cyclicity in nonpremenstrual phases may help explain why some women believe they have PMS but show no matching premenstrual pattern. We first explored whether women who said they had PMS were experiencing a M cluster in a nonpremenstrual phase rather than during the premenstrual phase. The same criteria were applied to other phases of the menstrual cycle as was applied to the premenstrual phase to determine the presence of other "diagnosable M" episodes. Specifically, 75% of all items within the phase must be negative and, for the McFarlane and Williams clusters, there must be no other phase with a M cluster. The occurrence of diagnosable M clusters for each menstrual phase for each say group is presented in Table 18. Table 18 Distribution of diagnosable M clusters for Current-Say and Never-Say self-designations Phase of the Percentage of women with diagnosable menstrual cycle 3-item M clusters Current-Say Never-Say Total (n=35) (n=23) (n=58) Menstrual 9 13 11 Postmenstrual 0 9 3 Midcycle 9 9 8 Post-midcycle 6 9 8 Premenstrual 14 17 16 Any phase 34 52 41 Percentage of women with diagnosable 5-item M clusters Current-Say Never-Say Total (n=35) (n=23) (n=58) Menstrual 11 13 12 Postmenstrual 0 4 2 Midcycle 3 0 2 Post-midcycle 0 4 2 Premenstrual 9 9 9 Any phase 23 30 26 106 Twenty-four women in this sample (41%) had M clusters that met the diagnostic criteria of McFarlane and Williams (1994) for some phase of the menstrual cycle. Nine of these women (37%) had a M in the premenstrual phase. Thus, approximately two-thirds (62%) had a M that could have resulted in a nonpremenstrual phase being diagnosed if the classification were available. For example, seven of these women (26%) could have been classified as having a Menstrual Phase Syndrome and five (21%) could have been classified as having each of a Midcycle Phase Syndrome or Post-midcycle Phase Syndrome. Thus, in this sample, M clusters that were exclusive to a particular menstrual phase and systematic over several menstrual cycles commonly occurred. Overall, more than 40% of the women demonstrated such a phenomenon, but only 15% met PMS criteria. The remaining proportion had other potentially diagnosable nonpremenstrual phases. A chi-square test of multinomial probabilities (McClave & Benson, 1982) indicated no significant difference among menstrual cycle phases in the number of diagnosable M clusters (rj>.05). These data support McFarlane and Williams' contention that an exclusive focus on the premenstrual phase as particularly troublesome may not be tenable. Such a position does not deny that for some women in this study, the premenstrual phase was characterized by unusual negative cyclicity. But only about half the women for whom this was the case labelled it PMS. Only five (9%) women both had predominantly negative premenstrual cyclicity and believed that they had PMS. Women from all say groups experienced nonpremenstrual M clusters that were distributed across the phases. Diagnosable M patterns occurred for all groups. A z-test of independent proportions indicated no significant differences between Current-Say and Never-Say women for any menstrual phase (p>.05). Chi-square analyses found no signficant links between the 3 say groups and the 107 distribution of diagnosable M clusters across the menstrual phases (see Table 19). These results do not support the hypothesis that women are simply mislabelling a negative menstrually-related pattern occurring in a nonpremenstrual phase. About a quarter (26%) of the sample had diagnosable M clusters according to the DSM-IV criteria (see Tables 18 and 19). Although symptomatic clusters did not occur often enough to be considered normative, they were not uncommon, even with a strict 5-item criterion. Tests of independent proportions revealed no significant differences in the proportion of Current-Say and Never-Say women who had diagnosable 5-item M clusters for any of the menstrual phases. Of the diagnosable M clusters, 80% were in the paramenstruum, that is, 47% menstrually and 33% premenstrually. One fifth (21%) of the women had a 5-symptom M cluster paramenstrually. In contrast, only 2 women (3%) had a diagnosable M within a nonparamenstrual phase (i.e., postmenstrual, midcycle, post-midcycle). The comparison between menstrual and paramenstrual M clusters was found to be statistically significant, using a test of correlated proportions (Glass & Hopkins, 1996), x2(1)=4.57, p_< 05. Although these criteria support the conclusions of the 3-item criteria that the premenstrual phase was neither exclusively nor prevalently experienced as negative, they did indicate the greater incidence of negative symptomatology during the paramenstruum than during the middle of the cycle. They also indicate that this experience was not rare, even when a minimum of 5 symptoms was required. Women's use of PMS does not appear to be explainable as a mislabelling of a nonpremenstrual group of symptoms, even when the symptom cluster was relatively large (i.e., 5 items). It is possible that women saying that they had PMS experienced M clusters in nonpremenstrual phases, but that these M clusters were not exclusive to a particular phase. To investigate this possibility, we also constructed a table (see 108 Table 19 Distribution of diagnosable M clusters for Always-Say, Sometimes-Say, and Never-Say self-designations Phase of the Percentage of women with diagnosable 3-item menstrual cycle M clusters Always-Say Sometimes-Say Never-Say Total (n=16) (n=22) (n=23) (n=61) Menstrual 6 14 13 11 Postmenstrual 0 0 9 2 Midcycle 13 5 9 8 Post-midcycle 6 9 9 8 Premenstrual 6 18 17 15 Any phase 31 45 52 44 Percentage of women with diagnosable 5-•item M clusters Always-Say Sometimes-Say Never-Say Total (n=16) (n=22) (n=23) (n=61) Menstrual 13 14 13 11 Postmenstrual 0 0 4 2 Midcycle 0 5 0 2 Post-midcycle 0 0 4 2 Premenstrual 13 5 9 9 Any phase 19 18 30 23 109 Table 20) in which all M clusters were included, not just those cases in which a single M was present across phases. This second examination was only necessary for the McFarlane and Williams (1994) 3-item criterion since the only instances of multiple M clusters using the 5-item criterion occurred during the paramentruum, and this did not affect PDD classifications. There were 39 M clusters for the sample across menstrual phases. The proportions of Never-Say (56%) and Current-Say (57%) women who had M clusters in some menstrual phase did not significantly differ, using a z-test of independent proportions (p_>.05). Only 18% of the M clusters were experienced by the Always-Say women in contrast to 44% by the Sometimes-Say and 38% of the Never-Say groups (see Table 21). These differences in proportions were not statistically significant, however, using a chi-square test. Women's use of "PMS" does not appear to be due to the greater occurrence of diagnosable and nondiagnosable M clusters, either premenstrually or during a nonpremenstrual phase. Finally, we explored whether women who said that they had PMS differed from those who said they did not in the distribution of T (Terrific) clusters (See Table 22). There were no cases in which T clusters occurred in more than one menstrual phase for any individual woman, using either the 3-item or a 5-item criterion. Over a third of the women had a T for one menstrual phase. If women were characterized as being unusually positive in relation to menstrual cycle phases, 36% of our sample would sustain this characterization using the McFarlane and Williams (1994) criteria. This percentage did not significantly differ (p>.05) on a test of correlated proportions from the percentage of women who had diagnosable M clusters in a single menstrual phase (41%), or who had M clusters when "exclusionary" criteria were removed (57%). Significantly more Never-Say women (22%) than Current-Say women (3%) had a T during the postmenstrual phase of their cycles, z=-2.3, p_<05. 110 Table 20 Distribution of diagnosable and nondiagnosable M clusters for Current-Say and Never-Say self-designations Phase of the Percentage of women with 3-item M clusters menstrual cycle Current-Say Never-Say (n=35) (n=23) Menstrual 20 17 Postmenstrual 6 9 Midcycle 11 13 Post-midcycle 14 9 Premenstrual 20 17 Any phase 57 57 111 Table 21 Distribution of diagnosable and nondiagnosable M clusters for Always-Say, Sometimes-Say, and Never-Say self-designations Phase of the Percentage of women with 3-item menstrual cycle M clusters Always-Say Sometimes-Say Never-Say Total 3) (n=61) 20 7 11 13 16 56 (n=16) (n=22) (n Menstrual 13 27 17 Postmenstrual 0 9 9 Midcycle 13 9 13 Post-midcycle 6 23 9 Premenstrual 13 23 17 Any phase 38 68 57 112 Table 22 Distribution of diagnosable T clusters for Current-Say and Never-Say self-designations Phase of the Percentage of women with diagnosable 3-item menstrual cycle T clusters Current-Say Never-Say Total (n=35) (n=23) (n=58) Menstrual 3 9 5 Postmenstrual 3 22 11 Midcycle 22 4 9 Post-midcycle 9 4 7 Premenstrual 3 4 3 Any phase 31 43 36 Percentage of women with diagnosable 5-item T clusters Current-Say Never-Say Total (n=35) (n=23) (n=58) Menstrual 0 0 0 Postmenstrual 0 0 0 Midcycle 11 0 11 Post-midcycle 0 0 0 Premenstrual 0 0 0 Any phase 11 0 11 113 The difference between these groups was not significant for any other menstrual phase. There was no significant relationship between the 3 say groups and the proportion of T clusters in each menstrual phase using a chi-square analysis (j)>05) (see Table 23). Only four women had 5-item T clusters. All had their T clusters in the midcycle phase and all were using the PMS term to describe their current menstrual experiences. Two were in the Always-Say group and two were in the Sometimes-Say group. None of the Never-Say group or women using PMS to describe only past experiences had a T associated with the menstrual cycle. Were women's self-designations related to the number of daily variables they experienced over the menstrual cycle? It was possible that, rather than differing in cluster distribution, women saying they did and did not have PMS differed in the number of significant prospective variables occurring in a particular menstrual phase. One plausible hypothesis was that the women who said that they had PMS had more significant items occurring premenstrually, but that there were too few to form a cluster. Moreover, if these items consisted of more than one symptom type, then this pattern would not have been detected by the previously presented parametric analyses. To test this hypothesis, a 5 (menstrual phase) x 2 (negative, positive cyclicity) x 2 (Current-Say, Never-Say) MANOVA was used. A significant valence (negative, positive cyclicity) main effect, F(1,56)=11.38, p.< 005, and say group by menstrual phase interaction, F(4,224)=2.71, p_<05, were found. In general, the mean for negative items (M=1-16) was higher than that for positive items (JVK81) (see Table 24). This result did not differ for the two say groups. Simple main effect analyses of the interaction revealed two statistically significant results. Women who said that they currently had PMS had more 114 Table 23 Distribution of diagnosable Tclusters for Always-Say, Sometimes-Say, and Never-Say self-designations Phase of the Percentage of women with diagnosable 3-item menstrual cycle T clusters Always-Say Sometimes-Say Never-Say Total (n=16) (n=22) (n=23) (n=61) Menstrual 0 5 9 5 Postmenstrual 0 9 22 11 Midcycle 13 14 4 10 Post-midcycle 19 0 4 7 Premenstrual 0 5 4 3 Any phase 31 27 43 34 Percentage of women with diagnosable 5-item T clusters Always-Say Sometimes-Say Never-Say Total (n=16) (n=22) (n=23) (n=61 Menstrual 0 0 0 0 Postmenstrual 0 0 0 0 Midcycle 13 9 0 0 Post-midcycle 0 0 0 0 Premenstrual 0 0 0 0 Any phase 13 9 0 0 115 Table 24 Mean number of uncharacteristically cyclical variables in each menstrual phase for Current-Say and Never-Say self-designations Mean number of uncharacteristically negative variables Menstrual phase Current-Say Never-Say Total Sample (n=35) (n=23) (n=58) Menstrual 1.60 1.52 1.57 Postmenstrual 0.60 1.26 0.86 Midcycle 1.14 0.74 0.98 Post-midcycle 1.20 0.87 1.07 Premenstrual 1.23 1.35 1.28 Across phases 5.77 5.74 5.76 Mean number of uncharacteristically positive variables Menstrual phase Current-Say Never-Say Total Sample (n=35) (n=23) (n=58) Menstrual 0.69 1.00 0.81 Postmenstrual 0.61 1.17 0.83 Midcycle 1.40 0.61 1.09 Post-midcycle 0.83 0.52 0.71 Premenstrual 0.63 0.61 0.62 Across phases 4.16 3.91 4.06 116 significant variables occurring during the midcycle phase (M=1 -29), F(1,56)=5.71, rj<05, and fewer during the postmenstrual phase (M=-60), F(1,56)=5.05, rj<05, than did women who said that they had never had PMS (M=-67 and M=1 -17, respectively). When the same analyses were re-run using 3 say groups, no significant group differences were found (see Table 25). The hypothesis that Current-Say women might experience more negative items premenstrually was not supported. Rather, self-designation was related to more midcycle changes and fewer postmenstrual changes, relative to Never-Say women. These changes were both positive and negative, since no valence interaction was found. Menstrual patterns in the number of items experienced as uncharacteristically high or low differed only with the use or non-use of "PMS" and not with the consistency of its use. The means were indeed below the 3-item and 5-item criteria of the clusters, thereby providing some additional information about which phases were most cyclical for each say group at the level of the individual variable. Were women's self-designations related to a more sensitive detection of cyclicity? It may be that women's criteria for menstrually-related change is more liberal, or sensitive, than that of researchers. Perhaps a single variable occurring uncharacteristically in a particular menstrual phase is sufficient to be detected by women as a pattern. If so, women who say they have PMS might be more likely to have at least one significant variable occurring negatively premenstrually and/or positively midcycle than women who say they do not. The proportion of women who had a minimum of one significant variable is shown for each menstrual phase in Table 26. A 5 (menstrual phase) x 2 (positive, negative cyclicity) x 2 (Current-Say, Never-Say) MANOVA revealed a significant main effect of valence (positive, negative cyclicity), F(1,56)=4.09, rj<05, and say group by phase interaction, 117 Table 25 Mean number of uncharacteristically cyclical variables in each menstrual phase for Always-Say, Sometimes-Say, and Never-Say self-designations Mean number of uncharacteristically negative variables Menstrual phase Always-Say Sometimes-Say Never-Say Total Sample (n=16) (n=22) (n=23) (n=61) Menstrual 1.81 1.55 1.52 1.61 Postmenstrual 0.63 0.55 1.26 0.84 Midcycle 1.31 1.00 0.74 0.98 Post-midcycle 1.00 1.31 0.87 1.07 Premenstrual 1.06 1.27 1.35 1.25 Across phases 5.81 5.68 5.74 Mean number of uncharacteristically positive variables Menstrual phase Always-Say Sometimes-Say Never-Say Total Sample (n=16) (n=22) (n=23) (n=61) Menstrual 0.56 0.77 1.00 0.80 Postmenstrual 0.69 0.55 0.61 0.88 Midcycle 1.69 1.05 0.74 1.05 Post-midcycle 1.13 0.55 0.52 0.53 Premenstrual 0.69 0.73 1.35 0.67 Across phases 4.75 3.80 3.91 118 Table 26 Distribution of cyclicity based on a 1-item liberal-criterion for Current-Say and Never-Say self-designations Percentage of women with at least one uncharacteristically negative variable Menstrual phase Current-Say Never-Say Total Sample (n=35) (n=23) (n=58) Menstrual 54 65 59 Postmenstrual 26 65 41 Midcycle 60 30 48 Post-midcycle 66 48 59 Premenstrual 54 52 53 Any phase 100 100 100 Every phase 0 0 0 Percentage of women with at least one uncharacteristically positive variable Menstrual phase Current-Say Never-Say Total Sam (n=35) (n=23) (n=58) Menstrual 43 48 45 Postmenstrual 34 57 43 Midcycle 57 43 52 Post-midcycle 51 35 45 Premenstrual 46 35 41 Any phase 89 100 93 Every phase 6 0 3 119 F(4,224)=4.60, p_< 005. More women experience at least one negative variable (53%) than had at least one positive (45%), and this negative-positive discrepancy did not vary across menstrual phases or between groups. Followup analyses on the interaction indicated that more women who said they currently had PMS (60%) experienced a significant item during the midcycle phase than did women saying they had never had PMS (40%), t(56)=2.43, rj<.05. In contrast, fewer Current -Say (31%) than Never-Say (61%) women had a significant variable during the postmenstrual phase, t(56)=3.10, p_< 005. In addition, fewer Current-Say women had a significant item postmenstrually (31%) than during either the midcycle (60%) or post-midcycle (59%) phases, F(4,136)=3.69, rj<01. When these results were re-examined using 3 say groups, a more precise pattern emerged (see Table 27). A 5 (menstrual phase) x 2 (postive, negative cyclicity) x 3 (Always-Say, Sometimes-Say, Never-Say) MANOVA revealed a say group by phase interaction, F(8,232)=2.69, rj<01. Subsequent Oneway ANOVAs indicated that it was specifically women who said they had always had PMS (69%) who were more likely to have a significant item during the midcycle than was true of the Never-Say group (37%). The women who said that they had sometimes had PMS (32%) were less apt than the Never-Say women (61%) to experience a significant variable postmenstrually. These additional findings suggest that it would be an oversimplification to characterize a single group of women who say that they have PMS. Rather, there appear to be two groups of women in our sample saying that they have PMS who experience the menstrual cycle differently. The hypothesis that women who say they have PMS are more likely to show premenstrual cyclicity than women who say they do not was not supported in these analyses, which used the most liberal criterion of only 1 item. More women who felt that they had always had PMS did, however, have a significant variable during the 120 Table 27 Distribution of cyclicity based on a 1-item liberal criterion for Always-Say, Sometimes-Say, and Never-Say self-designations Percentage of women with at least one uncharacteristically negative variable Menstrual phase Always-Say Sometimes-Say Never-Say Total Sample (n=16) (n=22) (n=23) (n=61) Menstrual 63 55 65 61 Postmenstrual 31 23 65 41 Midcycle 69 55 30 49 Post-midcycle 63 73 48 61 Premenstrual 44 59 52 52 Any phase 82 100 100 95 Every phase 0 0 0 0 Percentage of women with at least one uncharacteristically positive variable Menstrual phase Always-Say Sometimes-Say Never-Say Total Sample (n=16) (n=22) (n=23) (n=61) Menstrual 38 45 48 44 Postmenstrual 38 36 57 44 Midcycle 69 45 43 51 Post-midcycle 63 41 35 44 Premenstrual 44 55 52 44 Any phase 100 100 100 100 Every phase 6 5 0 8 121 midcycle that occurred at uncharacteristic levels for them over several menstrual cycles. But the midcycle item or items for this group did not fluctuate positively more than negatively, as would be expected on the basis of the "classic" menstrual pattern. Summary of T, M, and Cyclicity Findings Normative group analyses of women's prospective data showed no distinctive patterns for symptom types for women believing that they currently had PMS. To explore further women's use of and understanding of "PMS", the idiographic patterns of the women's data were examined more closely. Nine women (15%) were designated as having a "PMS" pattern based on the cluster of 3-symptoms criterion used by McFarlane and Williams (1994). This researcher-designation did not match women's own designations of PMS. Of the nine women who actually met 3-symptom researcher-criteria for PMS, only 1 said she had always had it, 4 said they had sometimes had it, and 4 said they had never had it. Moreover, 85% of the women who believed that they currently had PMS did not meet researcher criteria. Five women (9%) were designated as having a "PDD" pattern, based on the 5-symptom criteria defined in the DSM-IV. This researcher-designation also did not match women's self-designations. Three PDD women said that they currently had PMS and two said that they had never had it. We also examined the number of T and M cyclical clusters, the incidence of diagnosable M clusters (only one menstrual phase with an M cluster), and the total number of M clusters using both the 3-item and 5-item criteria (See Table 28 for a summary of statistically significant patterns). Using the 3-item McFarlane and Williams (1994) criterion, cyclicity was normative. The number of T and M clusters did not differ between the Current-Say and Never-Say women. Approximately 80% 122 Table 28 A summary of significant findings for statistical comparisons conducted on the idiographic data for Current-Say and Never-Say women Group Comparison Analysis Direction of Effect Across Groups Percent of women with 5-item M clusters Paramenstrual > Nonparamenstrual Across Groups Mean number of items found significant in prediction analyses Mean number of negative items > Mean number of positive items 2 Say Groups Percent of women with 3-item T clusters during the postmenstrual phase Never-Say > Current-Say 2 Say Groups Percent of women with at least one significant item during the postmenstrual phase Never-Say > Current-Say 2 Say Groups Percent of women with at least one significant item during the midcycle phase Current-Say > Never-Say 2 Say Groups Women with 5-item T clusters All women who had 5-item T clusters said that they currently had P M S and all T clusters occurred during the midcycle phase 123 of Sometimes-Say and Never-Say women had some T or M cyclicity, but only 56% of the Always-Say group did so. Thus, there were fewer researcher-based confirmations of the "PMS" designation as well as fewer T or M cyclical clusters for women who said that they had always had PMS than there were for women who used "PMS" intermittently or had never used the label. Using the 5-item DSM-IV criteria, cyclicity was not normative. Slightly more than a quarter (29%) of the women had a 5-item T or M cluster in a menstrual phase, and these women were distributed across say groups. Over 40% (41%) of women had a diagnosable 3-item M in some phase of the menstrual cycle. The incidence of diagnosable M clusters did not significantly differ among menstrual phases. Thus, the premenstrual phase was not particularly more troublesome than were other phases. About a quarter (26%) of the women had a 5-item diagnosable M. The distribution of diagnosable M clusters did not differ for the say groups for either a 3-item or 5-item criterion. Thus, women's self-designations could not be explained as either an accurate portrayal of a negative premenstrual M or as a mislabelling of a negative nonpremenstrual M, in the case where no other phase had a M. An examination of the 5-item M clusters revealed that 80% occurred during the paramenstruum. Significantly more women (21%) had 5-item M clusters during the paramenstruum than during the non-paramenstruum (3%), but this pattern was not related to women's self-designations. When the total number of 3-item M clusters was considered for each menstrual phase, without excluding those cases with a M in more than one phase, 39 M clusters were found. There were no significant differences in the distribution of these clusters for the say groups. Thus, for both those women having a single phase with a negative cluster and those women having more than one phase with a 124 negative cluster, the use of the "PMS" label was unrelated to any particular negative menstrual pattern. There was no significant difference between the proportion of women who had 3-item T clusters and the proportion who had either diagnosable M clusters or M clusters totalled without the exclusionary criterion. Significantly more women who believed that they had never had PMS had postmenstrual T clusters than women who believed that they currently had PMS. Three women had 5-item T clusters. All of these women had T clusters during the midcycle phase and all said that they had PMS currently. In sum, the women's self-designations did not generally match their patterns of negative cyclicity. Rather, women who believed that they have never had PMS were more likely to experience a cluster of 3 or more uncharacteristically positive items during the postmenstrual phase. With lower symptom criterion (3 items) for judging cyclical episodes, cyclicity was normal and the premenstrual phase was no more negatively experienced than any other phase. With higher symptom criterion (5 items), cyclicity was not normative and the paramenstrual phases were more negatively experienced than the non-paramenstrual phases. Women who said that they currently had PMS had more daily variables that significantly cycled during the midcycle phase and fewer during the postmenstrual phase than did women who said that they had never had PMS. Using a liberal criterion, more Current-Say women experienced at least one significant variable during the midcycle phase than did Never-Say women. Fewer Current-Say women had at least one significant variable during the postmenstrual phase than did Never-Say women. These latter two findings, however, were not universal across women who said that they had PMS. Analyses of the 3 say groups indicated that more women who believed that they had always had PMS had one or more cyclical items 125 during the midcycle phase than did Never-Say women. Fewer Sometimes-Say women than Never-Say women had one or more cyclical items postmenstrually. These findings did not vary by valence. This supports the conclusion that women saying that they had PMS did not demonstrate more negative cyclicity in any phase of the cycle. Thus, as more liberal criteria are used, differences related to women's self-designations become more evident, especially related to how consistently women use PMS. It may be that women's use of "PMS" is more related to levels of mild (1 item) or moderate (3 item) cyclicity than to more extreme (5-item) cyclicity. But none of the 3 sets of criteria matched women's self-designations in a manner fully consistent with a PMS pattern. Although women using "PMS" currently demonstrated a positive midcycle pattern, which matches the classic PMS pattern, they did not have a negative premenstrual pattern. It is possible, however, that, as McFarlane et al. (1988) conjectured, these women consider their positive midcycle experience to be normal or typical for them, so the preceding paramenstrual phases seem down or more negative by comparison. Analyses of Current-Say and Never-Say nonPMS Women The number of women meeting the researcher-designated criteria of PMS was not sufficiently large to investigate fully the interaction between self-designation and prospective PMS patterns. It is plausible that women who say that they have PMS and also meet researcher criteria differ from women who say that they have PMS and show no evidence of PMS. To explore this interaction partially, the single-measurement variables were analyzed for differences between the Current-Say and Never-Say groups among just the nonPMS women (n=49), that is, excluding the 9 women who met the researcher-designated criteria for PMS. The results of these analyses are reported in Tables 29, 30, and 31. The findings were generally very Table 29 Means and confidence intervals of single-measurement variables for nonPMS Current-Say and Never-Say women Variable Say Group Means Conf idence Intervals Menstrual Attitude Questionnaire Natural Current 27.00 24.91, 29.10 Never 21.84 19.25, 24.43 Predictable Current 29.10 27.07, 31.13 Never 21.84 19.34, 24.34 Bothersome Current 24.73 22.05, 27.41 Never 27.95 24.64, 31.26 Debilitating Current 47.37 42.86, 51.88 Never 37.05 34.29, 39.81 Should Ignore Menstruation Current 16.27 13.81, 18.73 Never 23.37 20.34, 26.40 Anger and Sadness Measures Anger at Home Current 10.50 9.20, 11.80 Never 8.72 7.12, 10.32 Anger Arousal Current 32.77 28.86, 36.68 Never 22.94 18.11, 27.77 Beck Depression Inventory Current 10.10 7.74, 12.46 Never 6.50 3.59, 9.41 Sadness at Home Current 9.63 8.30, 10.96 Never 9.39 7.75, 11.03 P s y c h o l o g i c a l Maltreatment of Women Inventory Domination-Isolation Current 28.96 26.19, 31.73 Never 29.67 26.25, 33.09 Emotional-Verbal Current 41.12 35.48, 46.76 Never 42.58 35.88, 49.28 Hassles and Uplifts Hassles Current .73 .59, .87 Never .49 .32, .66 Uplifts Current .75 .62, .88 Never .79 .64, .94 M a r i t a l Need Satisfaction Current 72.67 69.42, 75.92 Scale Never 73.23 70.44, 76.02 127 Table 30 Results of single-measurement variables for nonPMS Current-Say and Never-Say women Variable df Statistic p-level Effect S ize Power Partial T I 2 Menstrual Attitude 5,43 F=.34 P=.002 *** .95 Questionnaire Natural 1,47 F=8.11 P=.006 ** .84 .80 .15 Predictable 1,47 £=17.19 p=.ooo **** 1.22 .98 .27 Bothersome 1,47 F=1.93 p=.172 .41 .27 .04 Debilitating 1,47 E=7.04 p=.011 * .78 .74 .13 Should Ignore 1,47 E=11.24 p=.002 *** .98 .91 .19 Menstruation Anger and Sadness 4,43 F=2.28 p=.076 .62 Measures Anger at Home 1,46 F=2.44 p=.125 .47 .33 .05 Anger Arousal 1,46 F=8.20 p=.006 ** .85 .80 .15 Beck Depression 1,46 F=3.03 E=.088 .52 .40 .06 Inventory Sadness at Home 1,46 F=.04 p=.834 .06 .04 .00 (.001) P s y c h o l o g i c a l Maltreatment of 2,35 E=.00 (.002) p=.963 .06 Women Inventory Domination-Isolation 1,36 E=.06 p=.804 .09 .05 .00 (.002) Emotional-Verbal 1,36 F=.07 p_=800 .09 .05 .00 (.001) Hassles and Uplifts 2,46 F=2.50 p=.093 .48 Hassles 1,47 F=4.20 p=.046 * .60 .52 .08 Uplifts 1,47 E=.10 p=.756 .11 .05 .00 (.002) M a r i t a l Need 36 t=-.08 p=.934 .22 .45 Satisfaction Scale 128 Table 31 Percentages of nonPMS Current-Say and Never-Say women reporting abuse Variable Say Group Percentage Direction and p-level Reporting Abuse Strength of Association Physical Abuse by Current 20 phi=.28 e=.i05 a Romantic Partner Never 50 Sexual Abuse by a Current 63 phi=.16 p_=344 Romantic Partner Never 46 Emotional Abuse by Current 79 phi=. 16 P=. 329 a Romantic Partner Never 64 Physical Abuse Current 37 phi=.15 p=.296 in Adulthood Never 22 Sexual Abuse Current 43 phi=. 21 E=. 139 in Adulthood Never 22 Emotional Abuse Current 90 phi=.45 E=.002 *** in Adulthood Never 50 Physical Abuse Current 23 phi=.16 p=.259 in Childhood Never 11 Sexual Abuse Current 37 phi=. 23 p=.115 in Childhood Never 16 Emotional Abuse Current 63 phi=.36 E=. 012 in Childhood Never 26 129 similar to the 2-Say group results based on the entire sample of 58. There were only two exceptions. First, the Bothersome subscale of the MAQ was significant for the entire sample (p_=044), but did not approach significance for the nonPMS women. This discrepancy in findings may indicate an interaction between self-designation and a prospective PMS pattern. But power was notably lower for the nonPMS analysis (.27) than the analysis using 58 women (.58). Moreover, the means and confidence intervals changed very little with the removal of the PMS women. It seems that a loss of statistical power in dropping the PMS women from analysis is a more likely explanation for the different findings. Second, the Current-Say and Never-Say women differed significantly in their mean scores on the Beck Depression Inventory for the overall sample (n=58, £=.030), but not for the nonPMS sample (n=49, £=.088) . The £-level was close for these two analyses and the means (nonPMS analysis, M=10.10 Current-Say, M=6.50 Never-Say; whole sample, M=10.03 Current-Say, M=6.09 Never-Say) and confidence intervals were virtually identical. Power was low for the nonPMS analysis, but the effect size was moderate. Again, the difference in findings for the BDI seems to be an issue of power, rather than an indication of an interaction. Overall, it does not appear that the results differed in a meaningful way when women showing PMS patterns were excluded. This suggests that the differences found for the single-measurement variables were primarily attributable to self-designations, rather than to an interaction of self-designation and prospectively reported cyclicity, for nonPMS women. Not enough women met the researcher criteria, however, to analyze and generalize about possible differences between PMS women who say that they currently have PMS and those who say that they have never had PMS. Differences might be found between these two groups if enough such women could be found for a prospective study. This would be very 130 difficult, as only a small proportion of women (15.5% in this sample) recruited from the community who are unaware of the study's menstrual purpose will meet researcher-designated criteria for PMS. The proportion recruited from a PMS clinic would likely be higher, but then the women would not be naive to the menstrual purpose of the study. Causal Attributions of Women Who Said That They Have Experienced PMS How Did Women Who Said Thev Had PMS Explain Their Own Perceived PMS?. To investigate further the nature of women's self-designations, the women who said that they had PMS were asked what they believed caused their perceived PMS. It was possible that women's use or non-use of the label "PMS" was a less accurate indicator of women's menstrual attitudes and experiences than were their implicit theories about the origin of PMS. Women's open-ended responses to this question were coded into four categories: (1) the menstrual cycle as an independent variable (cause-iv), (2) the menstrual cycle as a dependent variable (cause-dv), (3) an interactive model of the menstrual cycle (cause-both), and (4) PMS as a mystery (cause-mystery) (See Method section for a more thorough description of these codes). There were 38 women who indicated that they had experienced PMS at some point in their lives and 36 of these women described causes for their perceived PMS. Two women's responses were not coded because they described the character of PMS rather than its cause(s). Nine women (25%) were coded as cause-iv, 8 as cause-dv (22%), 14 as cause-both (41%), and 3 as cause-mystery (8%). Because of its small size, the cause-mystery group was omitted from further analyses. Thus, analyses of causal attributions were conducted for 31 women. Only 131 this part of the sample is discussed in the following sections concerning women's explanations of their perceived PMS. Nearly half of the women who said that they had experienced PMS believed that it was caused by a combination of socio-environmental and physiologically-based factors (cause-both) (45%) (See Figure 13). The rest of the women were approximately evenly divided into those who subscribed to an independent variable or physiological model (29%) and those who held a dependent variable or socio-environmental model (26%). Women who treated the menstrual cycle as an independent variable provided a narrow and consistent range of causes that included references to hormones, chemicals, pregnancy, and nature. All of these responses treat the body as the origin of premenstrual symptoms. PMS is neither socially-induced nor socially constructed for these women. There was more variability in the causes supplied by the cause-dv and cause-both groups. Cause-dv women described such origins as life changes, exercise, sleep, travelling, poverty, marriage, moving, not making art, people doing stupid things, too many decisions, death of a parent, and a lack of social support. All of these attributions share the belief that the menstrual cycle can be personally or environmentally affected, and that this effect is responsible for the occurrence of PMS. Cause-both women discussed the co-occurrence of estrogen, hormones, vitamin B deficiency, chemical changes, aging, physiology, lymph nodes, and the CNS with stress, inactivity, eating, relationships, sex, diet, coffee, and weather. All but two of these women discussed how these factors combined to produce PMS. This type of variability reflects the many meanings that women may be communicating when using the term "PMS", especially women who are describing PMS as a dependent or interactive variable in their lives. 132 P e r c e n t 50 40 30 20. Cause-dv Cause-iv Cause-both Fig. 13. Causal Attributions for Women Who Say That They Have P M S 133 Women's stable (Always-Say) or intermittent (Sometimes-Say) use of "PMS" was examined for links to their beliefs about how their perceived PMS was caused, using a chi-square analysis. No significant relationship was found (p_=.323) (See Table 32). The women of this study did not appear to describe themselves as always having experienced PMS or as sometimes experiencing PMS based on their ideas about how their perceived PMS originated. Were Women's Explanations of Their Own PMS Related to Their Attitudes, Emotions, and Life Circumstances? There were no cause group differences for any of the single measurement variables. Women's attributions for their perceived PMS were not related to their menstrual attitudes, Beck depression scores, anger arousal, anger or sadness experienced at home, psychological maltreatment by a romantic partner, perceived hassles and uplifts, or marital need satisfaction. The means, confidence intervals, effect sizes, power, and partial n,2 for these results are in Tables 33 and 34. Overall, power was low for these analyses. The group sizes were small, so the lack of significant differences may represent Type II error in some cases. For example, some effect sizes were substantial. The effect sizes between the cause-iv (M=28.89) and cause-dv (M=22.13) groups was 1.18 and between the cause-dv (M= 2 2-13) and cause-both M=27.36) groups was .91 for the natural subscale of the MAQ. Statistical significance was nearly attained for this univariate test (p_=.053) and there was little overlap in the confidence intervals for the cause-dv group and the other two groups. Power was only .57. There is reason to suspect, therefore, that these differences are significant in the practical sense, but did not reach statistical significance because of Type II error. A reasonable hypothesis for future research is that women who use either a solely or partially biological explanation for their PMS are more likely, in general, to view menstruation as a natural part of 134 Table 32 Causal attributions provided by Always-Say and Sometimes-Say women Cause-dv Cause-iv Cause-both Always-Say 2 5 4 Sometimes-Say 6 4 10 135 Table 33 Means and confidence intervals for single-measurement variables for the cause-iv, cause-dv, and cause-both groups Variable Cause Group Mean Confidence Interval Menstrual Attitude Questionnaire Natural Cause-iv 28.89 24.97, 32.81 Cause-dv 22.13 17.97, 26.29 Cause-both 27.36 24.22, 30.50 Menstruation should Cause-iv 17.22 13.93, 20.52 be ignored Cause-dv 17.88 14.39, 21.38 Cause-both 14.71 12.07, 17.35 Bothersome Cause-iv 22.56 17.68, 27.44 Cause-dv 29.13 23.95, 34.31 Cause-both 24.71 20.80, 28.62 Predictable Cause-iv 29.89 26.33, 31.45 Cause-dv 28.88 26.17, 31.59 Cause-both 29.64 27.59, 31.69 Debilitating Cause-iv 41.56 34.01, 49.11 Cause-dv 46.63 38.62, 54.64 Cause-both 48.57 42.51, 54.63 Anger and Sadness Measures Anger at Home Cause-iv 10.67 8.00, 13.34 Cause-dv 10.75 7.92, 13.58 Cause-both 10.21 8.07, 12.35 Anger Arousal Cause-iv 33.44 25.58, 41.30 Cause-dv 34.00 25.66, 42.64 Cause-both 28.57 22.27, 34.87 Beck Depression Cause-iv 11.22 5.99, 16.45 Inventory Cause-dv 9.25 3.70, 14.80 Cause-both 9.14 4.94, 13.34 Sadness at Home Cause-iv 8.11 5.42, 10.80 Cause-dv 10.38 7.53, 13.23 Cause-both 10.36 8.21, 12.51 P s y c h o l o g i c a l Maltreatment of Women Inventory Domination-Isolation Cause-iv 31.33 25.99, 36.67 Cause-dv 33.20 27.53, 38.87 Cause-both 26.55 22.27, 30.83 Emotional-Verbal Cause-iv 41.56 29.36, 53.77 Cause-dv 44.60 31.64, 57.56 Cause-both 44.27 34.47, 54.07 136 Hassles and Uplifts Hassles Cause-iv .89 .58, 1.19 Cause-dv .91 .59, 1.24 Cause-both .66 .38, 0.87 Uplifts Cause-iv .77 .51, 1.03 Cause-dv .86 .58, 1.13 Cause-both .66 .45, 0.86 M a r i t a l Need Cause-iv 74.44 60.39, 88.49 Satisfaction Scale Cause-dv 67.40 52.50, 82.30 Cause-both 71.75 60.49, 83.01 137 •3-co CO CO _CD - O CO CO > d CD CO p CD U ) F CD i co rz CO CO 3^ CD CO o ro 2 o 0) "CD ~O ro £ i _ ro CL ~5 co CD CO tr CO 0-cu o CL CD N CO •s CD i t HI o .52 U—< CO -*—1 CO c o 52 ra, cf' o O _CD . Q CO CO > CD CM O o l 1^  o CD 3 CO CO C CD CD 3 § o CD » ^  -^ CN CD CO LO o al CD CN co LLl ^ > ° .> TD -O CD CD CD CO CO CO 3 3 3 co co co u o o oo o CO T -T - LO CD O CD CN al t i l oo CM CM ,^ > ° .> "O -O cb CD CD CO CO CO 3 3 3 CO CO CO o o o 3 o CO o "•+-» •o co CD 3 i _ O -4—1 CO ign Me be CM O CM CD CO CO LO al ill CD CD CD CO CO CO 3 3 3 CO CO CO o o o CD E o CO I CD sz -4—1 o CQ h- r- o CM O CM LO 00 ill -» > ° cb cb cb CO CO CO 3 3 3 co co co o o o _CD co •6 CD o LO CN N -CD T -LO LO CO al 1^ -o 1 1 1 ^ > o .> "O .Q cb cb cb co co co 3 3 3 CO CO CO O O O !5 CD Q O CD al oo CD h~ LLI CO CO CD c •o CO CO •Q c CO 3 S SS o o o o CD o CM CM O T -O CD al o CD O LLl -> > ° cb cb cb CO CO CO 3 3 3 CO CO CO O O O CD E o I •4—' co i CD a> c < LO o LO CN C -o t CO al o LLI >t1 CD CD CD co co co 3 3 3 CO CO CO o o o CD CD < co r-~- • CM CN O o 00 al oo CM CM" > ° CD CD CD CO CO CO 3 3 3 CO CO CO o o o c _g CO CO CD CD cT Q o c o 0 CD > CQ E CM q q co q LO q o 90' CM q CO q CM LO CO co q CM CD CM CD T— CO q (D S T -CM CN O CO 1— LO LO O T t CM CO CM CD CO LO CM T - T - o LO O LO O CD CO co o co CN CO LO LO CO T -ro t - CM O CO Ol r~ CO ol co -3-al CD CM h CD CM CD " i 61 CO CO LO Dl CN ol 00 ol CD CN CO ll'i ol F=.230 q T— II . LL.I F=1.82 F=1.63 F=.070 F_=787 F=1.50 F=.740 o CD 2,28 2,28 4,44 2,22 2,22 4,56 2,28 2,28 2,23 Cause-iv Cause-dv Cause-both Cause-iv Cause-dv Cause-both Cause-iv Cause-dv Cause-both Cause-iv Cause-dv Cause-both Cause-iv Cause-dv Cause-both Cause-iv Cause-dv Cause-both Beck Depression Inventory Sadness at Home Psychological Maltreatment of Women Inventory Domination-Isolation Emotional-Verbal Hassles and Uplifts Hassles Uplifts Marital Need Satisfaction Scale 139 womanhood than are women who ascribe their PMS to environmental influences. The effect size between the cause-dv (M=29.13) and cause-iv (M=22.56) women was strong (.92) for the Bothersome subscale of the MAQ, but this comparison did not attain statistical significance (rj=.175). Power was low (.35), however, so the test may have been unable reliably to detect a difference. Another hypothesis for future research with a larger sample is that women who believe that they have PMS and that it is socio-environmentally caused view menstruation as more bothersome than do women who believe that they have PMS that is biologically caused. Effect sizes were moderate or small for the other single-measurement variables, but with low power it is difficult to be confident that the moderate effect sizes are not potentially meaningful. The small effect sizes for Anger at Home, the Beck Depression Inventory, the Emotional-Verbal PMWI subscale, and the Marital Need Satisfaction scale were all quite small. All women from the 3 groups answered the questions about history of sexual, emotional, and physical abuse occurring in adulthood and in childhood. All of the cause-iv women answered the questions regarding history of abuse by a romantic partner, but only 4 (50%) of the cause-dv women and 11 (79%) of the cause-both women did so. All 4 cause-dv women and 7 of 9 (78%) cause-iv women reported that they had been sexually abused by a romantic partner. About half (46%) of the cause-both group reported sexual abuse by a current or past partner. Chi-square tests could not be used to test differences among the groups for the abuse questions because of the violation of assumptions of the expected frequency, so the statistical significance of the above proportions could not be established. Women from the 3 cause groups were evenly distributed in their reports on the other measures of abuse. The percentages of women reporting abuse for each cause group are shown in Table 35. 140 Table 35 Percentages of cause-iv, cause-dv, and cause-both women reporting history of abuse Variable Cause Group Percentage Reporting Abuse Physical Abuse by a Cause-iv 56 % Romantic Partner Cause-dv 50 Cause-both 55 Sexual Abuse by a Romantic Cause-iv 79 Partner Cause-dv 100 Cause-both 55 Emotional Abuse by a Cause-iv 89 Romantic Partner Cause-dv 75 Cause-both 73 Physical Abuse in Adulthood Cause-iv 38 Cause-dv 50 Cause-both 50 Sexual Abuse in Adulthood Cause-iv 38 Cause-dv 50 Cause-both 43 Emotional Abuse in Cause-iv 75 Adulthood Cause-dv 75 Cause-both 86 Physical Abuse in Childhood Cause-iv 22 Cause-dv 13 Cause-both 29 Sexual Abuse in Childhood Cause-iv 11 Cause-dv 25 Cause-both 36 Emotional Abuse in Cause-iv 56 Childhood Cause-dv 38 Cause-both 64 141 Were Women's Explanations of Their Own PMS Related to Their Prospective Patterns for Symptom Types? A 4 (symptom type) x 5 (menstrual phase) x 3 (cause-iv, cause-dv, cause-both) MANOVA revealed no significant differences in cyclicity among concentration and work performance, physical and activity, mood, and relationship symptom types in relation to the type of explanation that women gave for their perceived PMS. In other words, no evidence was found to indicate that women who differed in their beliefs about the origin of PMS significantly differed in the types of symptomatology that they experienced. Means, confidence intervals, and results for the MANOVA and univariate tests are in Tables 36 and 37. The tests of effects for these variables generally exhibited low power, but medium partial r | 2 effect sizes. There was higher power for the test of the 3-way interaction. An examination of the confidence intervals, however, showed a substantial overlap for all effects tested and p_-levels were high. An overall systematic pattern was not identifiable, either by the application of MANOVA or by inspection of means and confidence intervals. Were the Causal Explanations Provided By Women Who Said That They Had PMS Related To Researcher Designations? Four of the PMS-designated women and one of the PDD-designated women said that they had experienced PMS and provided an explanation for it. One of the PMS-designated women was coded as cause-iv and three were categorized as cause-both. The woman who met PDD criteria was coded as cause-both. These numbers are too small to draw conclusions concerning the relationship between women's theories about the causes of PMS and their likelihood of meeting diagnostic criteria for PMS or PDD. None of the women meeting researcher criteria 142 Table 36 Means and confidence intervals for prospective variables for the cause-iv,cause-dv, and cause-both women Effect Cell Mean Confidence Intervals Cause Group Cause-iv Cause-dv Cause-both 6.23 6.17 6.41 4.69, 7.77 4.54, 7.80 5.17, 7.65 Cause by Menstrual Phase Range of Means 6.11 to 6.53 Cause-iv: M ± .37 Cause-dv: M ± 39 Cause-both: M ± 29 Cause by Symptom Type Range of Means 5.62 to 6.87 Cause-iv: M + -80 Cause-dv: M + -85 Cause-both: M + - 5 5 Cause by Menstrual Phase by Symptom Type Range of Means 5.47 to 7.08 Cause-iv: M + - 1 9 Cause-dv: M ± 20 Cause-both: M + - 1 5 143 Table 37 Results of parametric analyses of prospective variables for cause-iv, cause-dv, and cause-both women Effect df Statistic p-level Power Partial T I 2 Cause 2,28 F=.66 p=.524 .15 .05 Cause by Symptom Type 6,54 F=.81 p=.568 .29 .08 Cause by Menstrual Phase 8,52 E=.52 p_=.84 .21 .07 Cause by Symptom Type by Menstrual Phase 22.19,310.61 F=.73 D=.808 .63 .05 144 provided a solely socio-environmental explanation for PMS in this study. Further research on this question would require a larger sample of PMS or PDD women. Were Women's Explanations of Their Own PMS Related to the Amount of Cyclicity They Experienced Over the Menstrual Cycle Phases? The previous analyses of women's T and M clusters showed little relationship between menstrual cyclicity and whether women said that they had PMS. The data were further explored to see whether women's T and M patterns varied with the ways in which those who said that they had experienced PMS explained its occurrence. About half the women in each group had a diagnosable 3-item M in a phase of the menstrual cycle. All but one of the diagnosable 3-item M clusters (7 of 8) for the cause-both women occurred during the paramenstruum. Half the diagnosable M clusters (2) in the cause-iv group occurred in the paramenstruum and half (2) were in the post-midcycle. For the cause-dv women, all of the diagnosable M clusters (4) were in the midcycle and post-midcycle phases. These numbers are too small to test for statistical significance, but it is interesting that for all of the women in this sample who attributed PMS to socio-environmental causes and who had a diagnosable M, it was the middle of the menstrual cycle that had a negative pattern. This did not occur for the women who used a biologically-based or combined model, whose diagnosable M patterns were only in the paramenstruum or post-midcycle. To complement this story, the percentage of women was tabulated in each cause group who had diagnosable 3-item T clusters. All but one of the cause-both women (7 of 8) having diagnosable T patterns experienced those patterns in the postmenstrual or midcycle phases. All diagnosable T clusters for the cause-iv women were in the postmenstrual (1), midcycle (2), or post-midcycle (1) phases. Only 1 cause-dv woman had a diagnosable T in the non-paramenstruum, 145 whereas 2 cause-dv women had diagnosable T clusters in the paramenstruum. Together, the diagnosable 3-item T and M cluster patterns suggest that about half the women in this sample who adopted a biologically-based model of PMS, either wholly (cause-iv) or partially (cause-both), experienced a cluster of negative variables in the paramenstruum and/or a cluster of positive variables in the non-paramenstrual phases. In contrast, about half the women who used a socio-environmental model of PMS (cause-dv) experienced a cluster of negative variables in the non-paramenstruum, and about a quarter had a cluster of positive variables in the paramenstruum. These patterns help us to understand better the perceptions of PMS among some of the women in this study, but they are based on too few women to be generalized. The occurrence of 5-item M and T clusters for women in the 3 cause groups was examined next. Only 6 of these women (19%) had a 5-item M in a menstrual phase and only 3 (10%) had a 5-item T in a menstrual phase. One woman in the cause-dv group had a M and this occurred in the post-midcycle phase. All 5 remaining M clusters were experienced by the cause-both group, four of which occurred in the menstrual phase and one of which occurred in the midcycle phase. No 5-item T cluster was experienced by the cause-dv women. Two cause-iv women and one cause-both woman had a T in the midcycle phase. There were only two patterns that were consistent across both the 3-item and stricter 5-item criteria. Four of the cause-both women (29%) had M clusters during the menstrual phase and 2 of the cause-iv women (22%) had T clusters during the midcycle phase. It seems that, using stricter criteria, there was little relationship between menstrual patterns and theories of PMS for women who said that they had PMS. The exception appears to have been a small group of women who said that they had PMS, who believed that it was caused by a combination of physiological 146 and environmental factors, and who experienced a reliable cluster of symptoms during the menstrual phase for several consecutive cycles. It was possible that women's explanations of PMS were related to the occurrence of M clusters, but that these M clusters were not exclusive to only one menstrual phase. We therefore tabulated all 3-item M clusters within each cause group, not just those cases in which a single M was present across phases. The distributions of M and T clusters for the cause-iv and cause-dv groups and T clusters for the cause-both group were the same as for diagnosable M and T clusters. Two additional cause-both women had a M cluster in a phase of the menstrual cycle that did not meet the exclusionary criteria of the diagnosable M clusters. These women had 4 M clusters and these occurred exclusively in the non-paramenstrual phases of the cycle. These cause-both women did not show the same type of pattern, then, as the cause-both women whose M clusters were restricted to a single phase. These women said that they had PMS, explained it as an interaction of the menstrual cycle and their environments, had more than one phase with a cluster of negative symptoms, and experienced this symptomatology in the non-paramenstrual phases of the cycle. A 5 (menstrual phase) x 2 (positive, negative cyclicity) x 3 (cause groups) MANOVA was used to determine whether the number of significant daily variables occurring in each menstrual phase differed for women who had different theories of PMS. No significant differences were found (p_>.05). A 5 (menstrual phase) x 2 (positive, negative cyclicity) x 3 (cause groups) MANOVA indicated a significant main effect of cause group, F(2,28)=4.32, p<.05, for the proportion of women who met a liberal criterion of at least one variable significant within any menstrual phase. Tukey post hoc tests, however, were not able to detect a reliable difference among the group means (p_>05). 147 Summary of Findings Associated with Women's Causal Explanations of PMS. Women who said that they had experienced PMS were asked to describe what they believed caused their PMS. Most of the responses (86%) could be coded into 3 categories. The first, cause-iv, was used to classify responses that treated the menstrual cycle as an independent variable that unidirectionally caused change for the woman and her environment during the premenstrual phase. This response was based on a medical or physiological model of menstrual cyclicity. The second category, cause-dv, included descriptions that characterized the menstrual cycle as a dependent variable in which change resulted from external stimuli. This response was based on a socio-environmental model of menstrual cyclicity. The third category, cause-both, was used for descriptions that combined characterizations of the menstrual cycle as both an independent and a dependent variable. These answers suggested that both physiological and environmental factors were operating, either mutually or additively, to elicit or exacerbate premenstrual changes. Nine women (29%) were coded as cause-iv, 8 (26%) as cause-dv, and 14 (45%) as cause-both. No significant relationship was found between the type of explanation women used and the consistency with which they described their perceived PMS histories (Always-Say, Sometimes-Say). Women who theorized differently about the PMS they believed they had did not signicantly differ in their reports of their romantic relationships, menstrual cyclicity, stress, anger, sadness, or depression. Statistical power was, however, low for these tests. Variables that demonstrated a moderate effect size and nearly attained significance included the Natural and Bothersome subscales of the MAQ. Type II error is especially suspect for these two measures, and re-examination of the relationship between women's causal attributions for PMS and their menstrual attitudes using a larger sample is recommended. Women from 148 the 3 cause groups were fairly evenly distributed in their reports of past emotional, physical, and sexual abuse. Women's cyclical patterns for 4 different symptom types - mood, concentration and work performance, physical and activity symptoms, and relationships - did not vary according to the way they explained their perceived PMS. None of the data analyzed across symptom types in relation to the cause groups indicated the classic PMS pattern for any group. Because power was relatively low for these analyses, means and confidence intervals were examined, but these also did not show any clear pattern of differences. Only 4 women who met PMS criteria and 1 woman who met PDD criteria said that they had had PMS and provided a causal explanation. None of these women ascribed their PMS solely to socio-environmental factors. Some interesting variations among the women of the 3 cause groups were revealed in an exploration of their idiographic cyclicity. The frequencies upon which these observations are based were not sufficiently large to be assessed for statistical significance and too small to be generalizable beyond this sample. The cause groups did not significantly differ, however, in the number of significant variables experienced for each menstrual phase or in the percentage of women who met a liberal 1-item criterion for each menstrual phase. Overall, women's implicit theories of PMS were not tied more clearly than were their self-designations to their menstrual cyclicity. A Review of the Findings Comparisons were made between the 35 women (60%) who reported that they currently had PMS and the 23 (40%) who reported that they had never had PMS. Summaries of the strength of the various findings are presented in Tables 38, 39, and 40. The average woman in this study who said that she currently had PMS was more likely to view menstruation as debilitating, natural, and predictable. She 149 Table 38 Statistically significant non-idiographic findings for which there was strong evidence Analyses 2 Say Group Comparisons Measure Menstrual Attitude Questionnaire Should be Ignored Debilitating Natural Predictable Anger and Sadness Measures Anger Arousal Emotional Abuse in Adulthood Prospective Data Symptom Type Symptom Type by Menstrual Phase Direction of Effect 3 S a y Group Comparisons Menstrual Attitude Questionnaire Should be Ignored Natural Predictable Anger and Arousal Measures Anger Arousal Never > Current Current > Never Current > Never Current > Never Current > Never Current > Never Direction of effect qualified by interaction Concentration and Work Performance: menstrual < other phases Mood: post-midcycle, premenstrual < postmenstrual premenstrual < menstrual < postmenstrual Physical and Activity: menstrual, post-midcycle < postmenstrual post-midcycle < premenstrual Across menstrual phases: Physical and Activity Variables < Concentration and Work Performance < Relationships < Mood Always-Say, Sometimes-Say > Never Always-Say, Sometimes-Say > Never Always-Say, Sometimes-Say > Never Always-Say, Sometimes-Say > Never 150 Table 39 Statistically significant non-idiographic findings for which there was relatively moderate or weak evidence Analyses 2 Say Group Comparisons Measure Menstrual Attitude Questionnaire Bothersome Anger and Sadness Measures Beck Depression Inventory Physical Abuse by a Romantic Partner Emotional Abuse in Childhood Menstrual Attitude Questionnaire Bothersome Debilitating Anger and Sadness Measures Beck Depression Inventory Emotional Abuse in Adulthood Direction of Effect 3 S a y Group Comparisons Current > Never Current > Never Current > Never Current > Never Never-Say > Always-Say Sometimes-Say > Never-Say Sometimes-Say > Never-Say Always-Say, Sometimes-Say > Never-Say 151 CO c o co TJ c CD E E o o CD DH ffl ja 'w co O CL CD o c co o i t "O c CD .2> tj CO CD g eg co .E CO T -TZ O co II c 3 CD i 03 CO CD L _ O O CO c CO CD > CD T3 JZ C CO CO -F 9 S E -rn — m •a CD CO CO ca E cz ZJ o - Q CD g X3 CD | 2 co 5 CD CO £ 2 8 -§ - C CO XJ 1$ 5 •o CO <=• CO O O "S Hi ™ ts iS CD co ^ CD LL £ 0> CO CO .9 c: i t CO cz > .?? c CO 3 •o CD .= cz co — CO CD £ II i CD CD « cz != co CO O -— *t p t CO _gj _a 'co CO o CL c o t3 2 ._ •6 2 CD O CD I— E £ — CO Q . £ i S CD fc- ' CD C o 8 ° o CD * £ •S o LU O co c CO -*—' CO . O o CO .co -*—' cz CO to ZJ CO CD W c ° 5 CD CO •D O ~ C O C C O co <r 2 o CD to i CD •D O E CD CO u . CD T J o E C O CD t-ZJ C O CO o CD ZJ 2 CD > CO C/3 CN to CO CD -§ CO CO CD CD E o I co i CD CO CZ < c o to .CO to CO T J CD CD | 8 § CO c CD E I o 21 O CO o - C CD -fe C3 ^ c^  s L ^ CD to TO > CL § £ co 1 CD > to c o o E LU 152 C O c o ' CO ' c CD E E o o CD ICC CD IS C O C O o la. CD o c co a5 -25 - CO o -a O CD c co o t= fc co 0 5 >. ~ — CO 'CO » CD a. T3 >-S l~" O CD si X CO CD CO r- ° ^ § CO -5 o tj ,,' CD 2 £ CD -C o = CO O CD 2 co > , p .a i- =5 •2>S S co CO £ •D co % ® 8- £ o -'co Q . M - CO S- ol 58 CO CO o t J co I— cu •4—' CD .£r I E co o ° - S i c E .o) > CO CO _co CO CD l . a l : CD o CD Q-l c "2 — a) ^ o3 1< o > LLI O O CO CO .O CO o CD iB = CD CO 1 "5 5 ' ~ C O <= C O 2 o CO l l ' o if CL CO 2 CD •D o E CO CD CO CO CD CO ^ CO ^ CD CO CO CO m —s n 3 0. CD co co CO 3 g 3 X fc - O CD O <a cj) tr CO >< .O CD CD co c 3 -£ •5 ro CO o ° £ CO g E E o L U cr: co ± i 0- < o o CO ;o x: £ CL O co n 3 1 CD -C CO O CD CL E E CD O O co -*—» -4—' CO CL CL jz E E Q. CO CO § >> >s iz co co cp. CO CO 2 153 C O c o CD E E 8 l a CD CO CO o ID. •JS co" c CD I ® £ CO ^ 'co co o co . i §• CD 2 CD 0 3 =" CO co co gcM o ro ro E CD L _ 3 CO ro CD E co >H -C X 3 .551 S CD CO £ . 1 CO . CD .2 •£ n — > CD •* O ° c ~ l -g .col 03 c CD 1< o > LU O O ro CD -•—1 CO l_ CD 1 o E ro -t—> c ro "co J 3 3 CO CO "c CO CO n 3 CO (Z CO , "co CO ro "£ ro CO t s CD !»= LU CD CO •D O B ° ~ to C CO • i --> o CD ro E CO ro E CO ro E CO C O CD I— 3 C A CO CD a 3 O lo CD 3 =4 QU co Ico CO CO CO C CD s a T - CM co ro co, ro < CO CO CD •g CO CO 1 8 C J CD CD E o x ro \ CD O) c < co, ro 5 CD ll CO CD E CD E o CO co, ro co CD E -t—' CD E o c o CO ro <: , >J CD CM co CD . cn • c < : o CD ICQ 154 Possible Recommendations nearly attained signficance (p_=.056), possible Type II error Extent to which Confidence Intervals Overlap substantial moderate slight substantial slight substantial substantial slight moderate substantial substantial substantial substantial substantial substantial substantial substantial substantial Magnitude of Effect or Association small moderate moderate moderate strong small small small small small small small small small small small small small Measures Sadness at Home 1) Never v. Always 2) Never v. Sometimes 3) Always v. Sometimes Hassles and Uplifts Hassles 1) Never v. Always 2) Never v. Sometimes 3) Always v. Sometimes Uplifts 1) Never v. Always 2) Never v. Sometimes 3) Always v. Sometimes Marital Need Satisfaction Scale 1) Never v. Always 2) Never v. Sometimes 3) Always v. Never Psychological Maltreatment of Women Inventory Domination-Isolation 1) Never v. Always 2) Never v. Sometimes 3) Always v. Never Emotional-Verbal 1) Never v. Always 2) Never v. Sometimes 3) Always v. Never 155 CO c q ' 1 CD E E o o , ® leu CD C O C O CD o = TO CD O Q -f ^ O ) CD CO X3 "D 'co CD CO c o TO <=-CD CO ~ ° ^ CD ,| 2 1^ >> 59 CD CD co 2 ro CD CD • D 3 CD k_ -*—< CO c CD fc p 2 CD o ,>> CD c c o CD CO c^  Ol w co CD IS I 3 o CD O c CD Q. • K -g i 5 ® c ai t< o > LLI O O CD "co CO "5 CD LLI CD CD c co a> < CD T o CO CD l_ 3 CO co CD CD co , 3 he co > > 1_ X3 a> CD C co tr 3 CO < o CD V a. ct: CO cc co >, CD CD CO c 3 - t < Q. co o co E o L U C C co > . 3 a. < CO 2 x: !c Q_ O • D o o CD SZ co o b 2 E Z LU O CD E o "S.I C O CD CO co .c CL — CO CL 5 P c E CD O 2 & ^ E £ 3 > >>co co C O .5 S co 03 cn cz CO CO - T - J o g> CO 03 o CD z: 5 CD E > CD CO cz O CO co C L E o o CD CD CO CD Pi 156 reported more frequent and enduring experiences of anger and was more apt to report that she was emotionally abused as a child. Weaker but still statistically significant findings indicated that this woman scored higher (but subclinically) on depression and was more likely to report having been emotionally abused during adulthood and physically abused by a romantic partner. Although not statistically signficant at the multivariate level, there was also evidence to suggest that the woman currently labelling herself with PMS might experience more frequent and intense hassles in her life. Prediction analyses revealed more mean variables that cycled outside this woman's normal range during the midcycle phase and a greater likelihood of having at least one significant variable during the midcycle phase than was true of women who said that they had never had PMS. The average woman in this dissertation who said that she had never had PMS was more likely to report that the effects of menstruation should be ignored. A weaker finding indicated that she viewed menstruation as more bothersome. The prospective data for the average woman who said she did not have PMS demonstrated more mean significant variables and a greater likelihood of at least one significant variable during the postmenstrual phase than was true of women who said that they currently had PMS. More of these women had a cluster of 3 unusually positive items that reliably occurred during the postmenstrual phase of several menstrual cycles. No significant differences were found between women who said that they had PMS currently and women who said that they had never had it for marital satisfaction, reports of sexual abuse as a child, reports of physical abuse in adulthood or childhood, uplifts, or psychological abuse by a romantic partner. The correlation between women's reports of sexual abuse in adulthood and of PMS was moderate (phi=.40), but did not quite reach statistical significance (p_=.057). This 157 might be an important variable for further research. Women's prospective cyclicity was explored for 4 different types of symptoms (mood, concentration and work performance, physical and activity, relationships) but no differences were found between the two groups. In order to shed further light on women's self-designations, the consistency with which the women used "PMS" to describe their menstrual histories was considered. Of the 61 women in the sample (the 58 women previously discussed plus 3 additional women), 26% reported that they had always had PMS (Always-Say), 36% reported having PMS intermittently since menarche (Sometimes-Say), and 38% reported never having had PMS (Never-Say). Re-analyses using 3 say groups indicated that women who said that they had always had PMS rated menstruation as more bothersome than did women who believed that they had never had PMS. Although it did not attain statistical signficance, a medium effect size was found between the Always-Say and Sometimes-Say groups for this subscale. There was some evidence to suggest that this non-significance may have been the result of insufficient power. These results suggest that women's views of menstruation as bothersome might be related to the consistency with which women use PMS. In addition, Sometimes-Say women had higher depression scores than did the Never-Say women. An examination of confidence intervals suggests that a difference might be found in future research on larger samples between women who say that they have always had PMS and those who say that they have sometimes had it. These results suggest that depression scores might be associated with how women are applying the label "PMS". In contrast, the Sometimes-Say women rated menstruation as more debilitating than did Never-Say women and a medium, but non-significant, effect 158 was found between Always-Say and Never-Say women. The direction of these findings supports those for the MAQ subscales that assess attitudes toward menstruation as predictable, natural, and something to ignore, in which differences were related to women's use or non-use of the "PMS" label, rather than how they used it. Findings for anger arousal and for reports of emotional abuse in adulthood also indicated that group differences occurred between women who believed that they did and did not have PMS. Women who said that they had had PMS were asked to explain what caused their PMS, enabling us to ask whether there were differences among women who used the label based on what they meant by its use. Four themes emerged in their responses: the menstrual cycle was believed to act as an independent variable and a physiological agent effecting premenstrual change (cause-iv), the menstrual cycle was seen as a dependent variable within which premenstrual changes were effected as a function of socio-environmental factors (cause-dv), the physiology of the menstrual cycle and socio-environmental influences were both believed to effect premenstrual changes (cause-both), the origins of PMS were unknown and mysterious (cause-mystery). Further analyses were conducted using 3 categories: cause-iv, cause-dv, cause-both. Amost half the women (45%) attributed their perceived PMS to a bidirectional model of physiology and environment (cause-both). The proportion of women who described only physiological (29%) or socio-environmental (26%) causes were roughly equivalent. Thus, consideration of socio-environmental origins by women in their theories of PMS was common (74%). Women's beliefs about whether they had always had PMS or whether they had sometimes had PMS were not related to their beliefs about what caused their PMS. In addition, women's implicit theories of their perceived PMS were not 159 significantly related to depression, sadness at home, anger at home, perceived stresses, marital satisfaction, reports of any type of abuse, or attitudes toward menstruation. Women who explained their PMS differently did not differ in their prospective cyclicity for 4 different types of symptoms (mood, concentration and work performance, physical and activity, relationships). They also did not differ in the number of variables that cycled abnormally or in the likelihood of experiencing at least one abnormally cycling variable for any menstrual phase. Overall, within the group of women who said that they had PMS, there were very few systematic differences found that related to the type of causal theories women were communicating in using "PMS". But, power was low for these analyses, so Type II error was a possibility. An examination of effect sizes and confidence intervals suggested that, in particular, future research might explore the relationship between women's causal attributions and their attitudes toward menstruation. Women's self-designations of PMS were compared with researcher-designations of PMS based on the 3-symptom method developed by McFarlane and Williams (1994) and PDD based on the 5-symptom DSM-IV criteria. Nine women (15%) met the criteria for PMS classification, of whom five said that they had PMS currently and four said that they had never had PMS. Five (9%) women met the criteria for PDD classification, of whom three said that they currently had PMS and two said that they had never had PMS. Of the women who said that they currently had PMS, 85% were designated as nonPMS and 91% were designated as nonPDD. Women's self-designations as not having PMS matched researcher classifications of nonPMS for 33% of the sample and nonPDD for 36% of the sample. But the data overall indicate that women's self-designations of PMS did not match researcher-designations, even with the measurement of 15 daily variables and the use of 2 different classification methods. 160 It seemed plausible that women were experiencing positive or negative cyclicity in other phases of their menstrual cycles and mislabelling it PMS, so analyses were conducted to see if women who said that they currently had PMS exhibited more 3-item or 5-item clusters of unusually positive (T) or negative (M) variables than did women who said that they had never had PMS: Significantly more of the women who reported no history of PMS than women who said that they currently had PMS had a 3-item T cluster during the postmenstrual phase. Only 4 women had 5-item T clusters, of whom all were in the Current-Say group and experienced the T cluster during the midcycle. The proportion of women having M clusters did not differ between the say groups for any of the menstrual phases. Most women had a 3-item T or M cluster in either one (41%) or two (29%) menstrual phases, suggesting that cyclicity was both common and normative. This was especially true for women who reported never having PMS (80%) and sometimes having PMS (80%). Just over half the women who reported always having PMS (56%) experienced M or T clusters. About a quarter (25%) of the sample had a 5-item T or M cluster and 88% of these women had only one cluster across menstrual phases. Thus, the use of a stricter criterion would lead to the conclusion that cyclicity was common but not normative. Over 40% (41 %) of the women had a 3-item M cluster that occurred exclusively in one menstrual phase. There were no significant differences among phases in the occurrence of M clusters. Recurring negative cyclicity on an organized cluster of 3 variables was common across phases. It was not a phenomenon either exclusive to or more frequent during the premenstrual phase. In contrast to the 3-item criterion, 80% of 5-item M clusters occurred in the paramenstruum (47% menstrually and 33% premenstrually). A fifth of the women sampled (21%) experienced a cluster of 5 unusually negative variables that recurred 161 during the paramenstruum over several cycles. Only 2 women (3%) had a 5-item M in the non-paramenstruum. This difference was statistically significant. Whereas the use of a 3-item criterion would inform the researcher that negative cyclicity was not more characteristic of any particular menstrual phase, the use of a stricter 5-item criterion would instead suggest that negative cyclicity was more pronounced paramenstrually. Neither of these cyclical patterns, however, was related to whether women said that they did or did not have PMS, or to the consistency with which women labelled themselves with PMS. It seemed possible that the women were very sensitive to fluctuations and that a more liberal criterion would better match their self-designations, so say group menstrual patterns were analyzed using a very liberal criterion of one variable that cycled significantly outside the woman's normal range. More women who said that they currently had PMS had at least one unusually cyclical variable during the midcycle phase than was true of women reporting no history of PMS. In contrast, more women who felt that they had never had PMS had one or more unusually cyclical variables during the postmenstrual phase than was true of women who felt that they had PMS. More specifically, the 3 say group analyses revealed that it was those who reported always having PMS who were more likely to have a cycling midcycle variable and those who reported sometimes having PMS who were less likely to have a cycling postmenstrual variable. The use of a 3-item criterion indicated that the occurrence of a T cluster in the postmenstrual phase was related to women's self-designations as never having had PMS. The distribution of M clusters was not related to women's use of "PMS". The use of a stricter 5-item criterion resulted in all T clusters occurring in the midcycle phase for women who believed that they had current PMS. The more liberal 1-item criterion suggested that more Current-Say women had cyclicity in the 162 midcycle and more Never-Say women had cyclicity in the postmenstrual phase. This liberal criterion was also able to detect which women who said that they had PMS exhibited these patterns. The greater midcycle cyclicity characterized Always-Say women and the lesser postmenstrual cyclicity characterized Sometimes-Say women. Moreover, no significant differences were found between the number of 3-item M and T clusters women experienced. Based on a 1-item criterion, more women experienced negative cyclicity than positive cyclicity. Taken together, these analyses suggest that more midcycle cyclicity is characteristic of women who say that they have PMS and more postmenstrual cyclicity is characteristic of women who say that they have never had PMS. They also suggest that women's use of "PMS" is not related to their negative cyclical patterns. Whether the use of the "PMS" label is related to positive cyclical patterns or more generally to the amount of cyclicity experienced, however, depends on the strictness of the criterion used to define "cyclicity". 163 Discussion Previous research has highlighted the discrepancy between retrospective and prospective measures of PMS (Magna & Hamilton, 1986; Boyle & Grant, 1992; Christensen & Oei, 1989; McFarlane, Martin, & Williams, 1988; McFarlane & Williams, 1994). These researchers have focussed on why so many women report PMS symptom patterns that are not confirmed prospectively. The findings of McFarlane and Williams (1994) indicate that women do not necessarily label the symptoms that they report as PMS when asked directly. The focus of this dissertation was on those same women's use of the label "PMS". Of particular interest was whether the women who said that they currently had PMS and the women who said that they had never had PMS differed in their life circumstances, attitudes toward menstruation, emotions, and prospective menstrual cyclicity. This study was an initial attempt to determine what women may be communicating about their lives when they say that they have PMS. Did Women's Self-designations Match Their Prospective Cyclicity? McFarlane and Williams (1994) reported very little relationship between women's self-designations of PMS and researcher-designations of PMS based on prospective data. In their report, researcher-designations were based predominantly on mood variables. It seemed plausible that women who said that they had PMS might be basing their self-designations on different symptomatology than mood. For this dissertation, women's cyclical patterns were examined for 15 variables that were categorized into 4 symptom types: concentration and work performance, physical and activity, mood, and relationship symptoms. Contrary to the hypothesis, women who believed that they had PMS did not differ in prospective 164 cyclical patterns for any symptom type from women who believed that they had never had PMS. The women were re-classified into PMS or nonPMS designations based on the criteria developed by McFarlane and Williams (1994), and into PDD or nonPDD designations based on the DSM-IV criteria for Premenstrual Dysphoric Disorder (APA, 1994). Nine participants met PMS criteria but only 5 of them said that they currently had PMS. Five women were classified as PDD and 3 of these women said that they had PMS. The findings using both of these criteria and 15 variables covering four symptom types were similar to those of McFarlane and Williams (1994) for 7 predominantly mood variables: only about half of the women meeting researcher-designated criteria said that they had PMS. Overall, 41% of women's self-designations matched their prospective patterns using both PMS and PDD criteria, but 79% and 88% of these matches were for women who said that they had never had PMS, and 88% were for women who showed no PMS or PDD pattern. In contrast, of the participants who reported having PMS, 86% did not meet PMS criteria and 91% did not meet PDD criteria. Self-designations of PMS matched researcher-designations of PMS and PDD for only 9% and 5% of the women, respectively. One implication of these findings is that researchers must be careful not to select "women with PMS" or control groups based on the woman's presentation to a clinic (e.g., Schnurr, 1988), symptom reporting during a gynecological checkup (e.g., Kuczmierczyk, Labrum, & Johnson, 1992), or attendance at PMS support groups and information sessions (e.g., Coughlin, 1990). These results also underscore McFarlane and Williams' (1994) caution against the recommendation of the DSM-IV criteria for diagnosis of Premenstrual Dysphoric Disorder that women's 165 self-designations of PMS be used in the interim until prospective data can be obtained. Do Women's Self-designations Reflect Cultural Stereotypes About PMS? One explanation for reports of PMS among women who do not meet prospective criteria has derived from attribution theory (Koeske, 1980a). Koeske (1986) has theorized that both positive and negative emotions are understood and labelled on the basis of situational cues and cognitive schemata, rather than on physiological responses. The cognitive schemas with which individuals interpret both illness (Woods, 1989) and emotion (Koeske, 1986) emerge out of past experiences and cultural expectations. Koeske (1983) has suggested that how women interpret recent emotional experiences may be linked to their retrospective awareness of the menstrual phase in which the experiences occurred. Evidence for Koeske's attributional approach has been provided by several studies (e.g., Klebanov & Jemmott, 1992; Koeske, 1980b, 1981; Koeske & Koeske, 1975; Ruble, 1977). This attribution theory would imply that a discrepancy between women's labelling of PMS and prospective evidence of PMS is not aberrant. Rather, given the same premenstrual experiences, women's differing self-designations should be related to their schemata about menstruation and to their environments. In this study, the women who reported current PMS and who reported no history of PMS did not differ in prospective evidence of premenstrual cyclicity, but did differ in their attitudes toward menstruation. Women who reported that they currently had PMS perceived menstruation to be more inevitable and more debilitating than did women who said that they had never had PMS. 166 Reviewers have documented the prevalence of Euro-North American stereotypes that characterize the menstrual cycle as universally negative and unhealthy (Laws, 1985; McFarlane & Williams, 1990; Parlee, 1975; Rodin, 1992; Ussher, 1991). Women's use of stereotypes to guide their recall of menstrual experiences may account in part for the stronger relationship found in this study between menstrual attitudes and self-designation as having PMS than between prospective cyclicity and self-designation (Brooks-Gunn, 1985; McFarlane, Martin, & Williams, 1988; McFarlane & Williams, 1994). This explanation is consistent with Koeske's schematic approach. In this study, it was found that women who said that they currently had PMS perceived menstruation as more debilitating and signified more by mood and body changes, whereas women who did not use the label "PMS" perceived menstruation as simply bothersome, the effects of which women should be able to ignore. It may be that women who believe that they have PMS generally endorse cultural stereotypes more strongly. But theorists have also cautioned that women's perceptions about their menstrual experiences are neither entirely negative nor simply reflections of mythology (Birke & Best, 1982; Chandra & Chaturvedi, 1990; Chrisler, Johnston, Champagne, & Preston, 1994; Koeske, 1980b; McFarlane & Williams, 1990, 1994). Indeed, in this study women who reported currently having PMS viewed menstruation as more natural than did women who reported no history of PMS. Women using the "PMS" label were more likely to endorse items that characterized menstruation as healthy, affirming, and a way to be aware of their bodies. These results support those of Brooks-Gunn and Ruble (1980), who reported that women attributed both positive and negative aspects to menstruation, and that women's theorizing about their menstrual cycles was not reducible to a unidimensional stereotype. Moreover, the women's self-designations did not 167 support the view that women's popular mythology perceives the premenstrual phase to be a consistent and universal problem. Only 26% of the women characterized the premenstrual phase as consistently troublesome (Always-Say), and 37% indicated that they did not find the premenstrual phase to be particularily troublesome at any time during their menstrual years (Never-Say). About a third (36%) of the women used "PMS" to describe an experience that occurred occasionally but not inevitably (Sometimes-Say). In other research, women's causal attributions of symptomatology have been found to influence their symptom reports. Fewer symptoms were reported by women led to believe that PMS is psychologically caused than by women informed that PMS is physiologically-based (Fradkin & Firestone, 1986). Participants in another study indicated that they would be more apt to seek medical consultation for symptoms deriving from a physiological origin than from a psychological origin (Robbins & Kirmayer, 1991). These findings suggest that women's use of PMS in characterizing the nature of their menstrual symptoms would be related to their beliefs about the cause of their perceived PMS. But contrary to this expectation, there were no significant differences in this study between the causal attributions of women who reported that they had always had PMS and those of women saying that they sometimes had PMS. Of the women who said that they had experienced PMS, 29% ascribed it to physiological causes, 26% ascribed it to socio-environmental factors, and 45% believed that it was a consequence of both. The distribution of these explanations supports the notion that women's theories about menstruation are varied and complex. Nearly half the sample subscribed to an interactive model similar to the one that has become increasingly accepted and supported by menstrual cycle researchers (Frieze, 1978; Koeske, 1986; Parlee, 1981; Sommer, 1981). Nearly 168 75% of the women who said they had PMS believed that psychosocial factors played at least a partial role in their PMS. Overall, the results support Koeske's theory that menstrual attitudes are related to how women label their menstrual experiences and her assertion that women's theories about menstruation are not simplistic. Nash and Chrisler (1995) recently found that women who read PDD criteria with a PDD heading were more likely to attribute symptoms to themselves and to other women than were women who read PDD criteria with a gender-neutral heading. To the extent that attitudes toward menstruation are related to women's conclusions that they are ill, professionals must act responsibly in the dissemination of information. Are Women's Self-designations Related to Stress? Koeske theorizes that situational cues play a role in the labelling of menstrual experiences. In this dissertation, almost three quarters (74%) of the women who said that they had PMS spontaneously mentioned stress as a contributing factor when asked what caused it. This result is consistent with the observation of Woods et al. (1994) that women have attributed perimenstrual symptomatology to stress. Many women in this study described the kinds of chronic stressors (e.g., financial hardship) and life events (e.g., job loss, death of a parent) that researchers have linked to the etiology of PMS (Clare, 1985; Futterman et al., 1992; Kerstner, 1986; Koeske, 1980a; Mitchell et al., 1994; Woods et al., 1994; Woods et al., 1995). Some resesarchers have noted that past abuse history may act as a chronic stressor (Miccio-Fonseca et al., 1990). In this study, more women who said that they currently had PMS reported having been emotionally abused as a child and physically abused by a romantic partner. Other researchers (Futterman et al., 992) have noted that relationships, in general, are the primary stressors in women's lives. For example, McDaniel (1988) and Reid (1985) have indicated that women who live 169 with men report more problems with PMS. In this dissertation significantly more women who said that they currently had PMS (83%) were sexually active with a man or men than was true of women who said that they had never had PMS (54%). The effects of stress occurring in relationships may not be phase-specific, but the attributions associated with the stress may be phase-specific. Stress-related physiological changes and experiences have been shown to be ascribed to situational factors when they occur intermenstrually, but to menstrual physiology when they occur during the menstrual or premenstrual phases (Koeske, 1975). Thus, these stressors may not affect prospective premenstrual reports but may affect women's recall of "having PMS". Other measures of situational factors used in this study did not attain statistical significance. These included hassles and uplifts, marital need satisfaction, psychological maltreatment by current romantic partners, sexual abuse history, physical abuse in childhood and adulthood, and emotional abuse in adulthood or by a partner. Marginally significant findings for two of these variables (hassles, having been sexually abused as an adult) were in the direction consistent with Koeske's model (higher for Current-Say women), indicating that they would be important to include in future research. Moreover, when outliers were excluded, Current-Say women did report significantly more hassles than did Never-Say women. This finding is representative of the majority of the women in the study, but is not representative of all of the sample. Overall, in this study, women's attitudes toward menstruation were related more to their self-designations regarding PMS than were the variables that were selected as possible situational cues. Contributions From Illness Cognition Research In light of the discrepancy between self-designations and researcher-designations regarding PMS, McFarlane and Williams (1994) raised the question of 170 who is most likely to seek and receive appropriate treatment for menstrually-related difficulties. In order to assess accurately and treat the menstrual concerns with which women present, professionals must understand what women are communicating when they say that they have "PMS". Whereas Koeske has theorized specifically about the menstrual cycle, other researchers have proposed similar models to explain how people label illness generally. In a study by Lau et al. (1989), participants showed an overwhelming inclination toward labelling symptoms together as an illness when asked to describe the last time that they were sick. Lau et al. found that the illness cognitions that these labels represented were relatively stable, organized schemata that were applied to many different diseases. Evidence for the roles of past experience and of sociocultural factors in the development of illness representations comes from research by Schwartz and Gramling (1994), who found that social learning variables accounted for significant variance in symptom reporting. The use of illness schemata may influence individuals to ascribe current symptoms to origins that are consistent with their past physical or mental health experiences (Robbins & Kirmayer, 1991). Klebanov and Jemmott (1992) have noted that PMS research and illness cognition research both provide evidence for a general tendency by individuals to interpret their internal states according to expectations. Both approaches have originated from Schachter and Singer's (1962) theory that physiological arousal is accorded an emotional label based on situational cues (Baumann et al., 1989; Cameron etal., 1995; Koeske, 1980a, 1980b, 1981, 1986; Koeske & Koeske, 1975). Both hypothesize that the physical sensations of stress may be either assigned an illness label or attributed to environmental factors, depending on the context (Cameron, et al., 1995; Koeske, 1986). The types of illness schemata used by 171 individuals have been related to patterns of self-care, physician utilization, and compliance with prescribed treatments (Robbins & Kirmayer, 1991). Some authors have further delineated the role of culture in the formation of illness schemata by theorizing that patients label themselves as ill or symptomatic when their body experiences deviate from cultural norms (Chrisman & Kleinman, 1983; Freidson, 1970; Stoller & Forster, 1994). The results of this study regarding attitudes toward menstruation suggest that women who use the "PMS" label differently also differ in what they perceive to be normative menstrual experiences. These differing representations are informative to practitioners because it is the interpretation of symptoms that has been found to be most related to the propensity to consult a physician (Stoller & Forster, 1994). Moreover, once women have labelled themselves or been labelled with PMS, they are more likely to behave in a manner consistent with their perception of the label (Baumann et al., 1989; Woods, 1989). Individuals who are assigned an illness label actively seek corresponding symptoms and individuals who experience symptoms actively search for a label that explains their occurrence (Baumann et al., 1989; Robbins & Kirmayer, 1991). The search for a match between symptoms and label implies that the two are mutually reinforcing. This can result in the inaccurate adoption of a label or in symptomatology that is resistant to disconfirmation. For example, Meyer et al. (1985) found that even though 80% of the adults in their study agreed that hypertension was asymptomatic, 88% believed that they could recognize when their own blood pressure was elevated. Furthermore, they identified specific symptoms that they believed signalled their episodes of high blood pressure. In this dissertation, 52% of the women used "PMS" to characterize their menstrual experiences even though their prospective data showed no evidence of such a pattern. 172 For the women in this study, no relationship was found between their use of the label "PMS" and their beliefs regarding the cause of PMS. McFarlane and Williams (1994) found relatively little relationship between the same women's retrospective symptom reports and their self-designations as having or not having PMS. Robbins and Kirmayer (1991) have explained the failure of attributions to predict illness behaviours by distinguishing between causal attributions of illness and causal attributions of symptoms. Indeed, Lau et al. (1989) found that it was symptom reports, not labels, that were most strongly related to the propensity to visit a physician. In PMS research, however, treatment-seeking has been shown to bear little relationship to actual prospective confirmation (Ekholm & Backstrom, 1994; Stout & Steege, 1985). Treatment-seeking for PMS has been found to be associated with a greater likelihood of current or past psychiatric disorder (Miccio-Fonseca et al., 1990; Paddison et al., 1990). Thus, it may be that women who present for treatment are communicating some form of symptomatology that is not necessarily menstrually-related. For example, in this study, women who said that they currently had PMS had higher depression scores and reported more frequent, intense episodes of anger. All of the six women with depression scores above the clinical cutoff (BDI score above 15) said that they had PMS. These findings are consistent with other studies showing greater depression among treatment-seekers (Christensen & Oei, 1995; Rivera-Tovar, Pilkouis, & Frank, 1992). Regardless of the accuracy of the label women assign, the symptoms are probably distressing (Robbins & Kirmayer, 1991). It is important, therefore, that clinicians attend to and identify the concerns of women who are saying that they have "PMS", even when PMS patterns are prospectively disconfirmed. Individuals respond differently to chronic symptoms than to acute symptoms (e.g., Woods, 1989). Although both Always-Say and Sometimes-Say women 173 labelled th