Feminism and the quality of life

DEBORAH C. POFF AND ALEX C. MICHALOS FEMINISM AND THE QUALITY OF LIFE (Received 17 October, 1986) ABSTRACT. A measure of feminism is introduced, and a case is made for the acceptability of its levels of reliability, criterion-related, content, construct and discriminant-validity. Feminism is shown to be related to such features of the quality of life as happiness and being a good person. Survey results are reported from a sample of 431 members of the Canadian Research Institute for the Advancement of Women and 413 undergraduate women from the University of Guelph. 1. INTRODUCTION The introduction of any new social scientific scale or measure requires the satisfaction of at least five conditions. In particular it must be shown that the scale or measure has acceptable levels of reliability, criterionrelated, content, construct and discriminant validity. In the following sections, we will introduce the Poff-Michalos Feminism scale (PMF scale for short) and show that it satisfies these five conditions. As suggested in the principles of a feminist epistemology articulated in Poff (1985), we believe that even the fairly technical task of constructing a scale of measurement for social scientific research is typically motivated by and has important consequences for a more or less explicit social and political agenda. In the present case, the development of a quantitative measure of feminism would provide a succinct account of a set of beliefs and attitudes that incompletely but fairly accurately characterize contemporary Canadian feminism. Such an account is necessary for pursuing certain kinds of feminist scientific research programs and for clarifying the social and political vision sustaining the feminist movement. There is considerable concern in the feminist movement today that it is an aging movement. Misconceptions among young women about the feminist agenda may be more easily corrected when we have a set of clear, empirically warranted statements of feminists' beliefs and concerns. It is also important to have such a set of statements to be able to give a relatively authoritative and collective Social Indicators Research 20 (1988) 445-472. © 1988 by Kluwer Academic Publishers. 446 DEBORAH c. POFF AND ALEX C. MICHALOS response to politically reactionary groups like Realistic, Equal and Active for Life (R.E.A.L.) Women when they mistakenly or deliberately misperceive and misrepresent feminism in Canada. In the next section (2) we briefly review earlier studies designed to develop similar scales, and we provide some of the motivation for our own investigations. Following that we give an overview of the main sample used for our empirical data and an account of our questionnaire (Section 3). Since some feminists are opposed to quantitative studies in principle, they create special problems for survey researchers. We address these issues in Section 4. In the fifth section we explain our treatment of missing values. Associations among various types of feminism are shown in Section 6. Sections 7 and 8 contain the heart of the paper. In the former, the structure of the PMF scale is explained. Besides making a case for the content-validity of the scale as a whole by examining its parts, a case is also made for its reliability by revealing its internal consistency as measured by Cronbach alpha coefficients for the whole scale and selected parts of it. In Section 8 a case is made for the criterion-related, construct and discriminant-validity of the PMF scale. The first two points are made by confirming the following hypotheses. H 1. Being a feminist as measured on the PMF scale is (statistically significantly) positively correlated with (a) considering oneself to be a feminist, (b) describing oneself as happy, (c) describing oneself as independent, (d) describing oneself as truthful, (e) describing qneself as compassionate, (f) describing oneself as friendly, (g) being a good person, (h) considering one's family and career as equally important, (i) having high self-esteem. H2. Being a feminist is negatively correlated with describing oneself as competitive. H3. Being a feminist is not correlated with describing oneself as being yielding, gullible, childlike, conventional, aggressive, conscientious, assertive, loyal, understanding, warm, willing to take a stand, individualistic, having leadership abilities or loving children. FEMINISM AND THE QUALITY OF LIFE 447 The discriminant validity of the PMF scale is shown by applying it to a convenience sample of 413 undergraduate women at the University of Guelph. The final section contains some brief concluding remarks. 2. EARLIER STUDIES One of the best earlier attempts to develop a quantitative measure of feminism was an unpublished manuscript produced at the Center for the Sociological Study of Women at the University of Oregon (Acker et ai., 1974). The authors constructed the scale "to measure feminist consciousness as expressed in the Women's Liberation Movement of the early 1970s. Consequently, it reflects the ideology of that period." The scale has 36 items which were selected from "a pool of approximately 350 statements constructed by active feminists on the basis of their experience in feminist groups and their reading of feminist literature." It has broad content-validity, including items on "family and household, work, sex life, relationships with women and with men, education, political action, beliefs about female-male personality differences, reactions to sexism, parenthood and child rearing, and selfconcept." Its criterion-validity was assessed against a 3-item index including the following: "How do you feel about Women's Liberation?" ''What organizations are you active in?" and "I have participated, in the last year, in an organization for women's rights." The correlation between the sum of 35 of their scale items and the 3-item index was 0.80. Similar results were found for samples of women (mainly undergraduates) drawn in 1972 (N == 264) and 1973 (N == 300). Our main problems with this scale concern its content validity. Although it is suitably broad in scope, many of the items do not appear to provide a clear idea of a feminist agenda or program. Worse than that, some seem to indicate a relatively trivial agenda. For example, consider the following items: "I shave my legs regularly." "When a man opens a door for a woman, this symbolizes women's status as weak and inferior." "I have or am planning to learn some form of physical selfdefense (such as judo or karate)." "If I had to choose, I would rather be a nurse than a doctor." Although it is true that in certain contexts or with certain analyses of the symbolic meaning of such things as leg shaving, door opening and so on it may be shown that these items are 448 DEBORAH C. POFF AND ALEX C. MICHALOS not trivial, they may appear trivial to people unfamiliar with the appropriate analyses. They frequently suggest misleading images of feminists that are readily picked up by the popular media and widely disseminated. Thus, we think it is wise to try to exclude such items from standard measures. Dempewolff (1974) introduced a scale designed to measure "attitudes toward the aims of the women's movement" because she thOUght that other measures designed for similar purposes were ''flawed by tendentious or confusing items, lack of adequate validation, or incomplete coverage of issues by items." She began her selection of items from the Kirkpatrick Belief-Pattern Scale for Measuring Attitudes toward Feminism (Kirkpatrick, 1936). Her 56-item scale and its two shorter forms A and B showed excellent internal consistency scores (Cronbach alpha coefficients) of over 0.90. The total set of items was not given in this publication, but a sample was given that included the following: "A woman could be just as competent as a man in a high political office." "Objections which one might have to the use of obscene language should bear no relation to the sex of the speaker." "A wife should take the lead and suggest sexual intercourse if she wishes it." The scale was reported as "demonstrating known-groups validation" by discriminating individuals belonging to groups known to favour or oppose the women's movement. Nielsen and Doyle (1975) distinguished feminists from nonfeminists on the basis of responses to the question: "Do you consider yourself a member of a women's liberation group or organization?" with the response categories "Yes", "No, but 1 would like to be" and "No". Feminists, they found;had "a higher regard for women in general and a slightly more positive self-concept." Smith and Self (1981) sorted 279 women students into "feminists" and "traditionalists" on the basis of responses to the question: "I consider myself a feminist (one who supports the goals of the women's liberation movement)." On the basis of responses to 21 items "pertinent to the area of changing sex-role attitudes" they concluded that "the primary distinction between the two groups appears to be that feminists are inclined to support equalitarian personal relationships, indicating a willingness to alter the current social status quo vis-a-vis men." Smith, Ferrez and Miller (1975) constructed a 20-item scale that FEMINISM AND THE QUALITY OF LIFE 449 they regarded as measuring "attitudes toward feminism", but which Singleton and Christiansen (1977) regarded as "a measure of prejudiced or authoritarian attitudes toward women." The original scale was shown to have an alpha coefficient of 0.91, and two shorter forms had coefficients above 0.80. The shortest form had the following five items with the highest factor loadings from the original scale. "It is all right for women to work but men will always be the basic breadwinners." "A woman should not expect to go to the same places or have the same freedom of action as a man." "Realistically speaking, most progress so far has been made by men and we can expect it to continue that way." "A woman should be expected to change her name when she marries." ''Women who join The Women's Movement are typically frustrated and unattractive people who feel they lose out by the current rules of society." The original scale had good discriminant-validity, clearly distinguishing members of groups favourable or opposed to the Women's Movement. Aslin (1977) identified ''feminist clinicians" among psychotherapists by taking names from "The Feminist Therapists Roster of the Association for Women in Psychology." Foster, Strudler-Wallston and Berger (1980) determined ''feminist orientation" on the basis of responses to the question: "Defining a feminist as someone who believes in total equality between males and females, do you consider yourself to be a feminist?" with the response categories "Definitely", "Somewhat", and "Definitely not". Wilson (1982) claimed that "since a search of the literature turned up no satisfactory measure of feminist attitudes a new questionnaire was designed for (his] study." He divided his sample of 809 readers of She magazine into "feminist" and "nonfeminist women" on the basis of their views about the ''funniness of two cartoons". One cartoon "was female assertive in content (a wife waiting in the hallway to punish a philandering husband with the sight of a newly purchased fur coat) and the other chauvinistic (a lift-lady about to be taken advantage of by a male passenger who had jammed the lift between floors with his umbrella). The assumption was that feminist women would be relatively less appreciative of the latter joke." Alspach (1982) used a 3-item index of "non-traditional sex-role attitudes", which probably captures some feminist views, to test a 450 DEBORAH C. POFF AND ALEX C. MICHALOS hypothesis suggested by Friedan (1963), namely, that "women with non-traditional sex-role attitudes are happier than those with traditional sex-role attitudes." The items were as follows: "Women should take care of running their homes and leave running the country up to the men." "Most men are better suited emotionally for politics than are most women." "If your party nominated a woman for president, would you vote for her if she were qualified for the job?" His hypothesis was not supported, but its opposite was. Capturing another feature of feminism, Beere et al. (1984) constructed a sex-role Equalitarian Scale containing 19 items, and showed that it had good internal consistency and that "women scored more equalitarian than men, ... psychology students scored more equalitarian than business students, [and) both student groups scored more equalitarian than senior citizens and police." Finally, Berryman-Fink and Verderber (1985) developed "a measure of college students' attributions of the term feminist". They wanted to "assess the evaluative connotation of the label feminist." Seven hundred and sixty-eight undergraduate students (361 males, 407 females) completed a questionnaire containing 91 semantic differential items ''relevant to the term feminist". The result was a 54-item scale in which the following characteristics (among others) were attributed to feminists. They were "seen as more logical, knowledgeable, realistic, intelligent, caring, flexible, comforting, good, fascinating, aggressive, extroverted, activist, more likely to be working outside the home, opinionated, forceful, ambitious, independent, career oriented, active, strong, noncomfonning, dominant, assertive, busy, talkative, energetic, for equal wages, liberation, equal rights, in support of NOW [National Organization of Women), women's lib and the ERA [U.S. Equal Rights Amendment], heterosexual/straight, and female." 3. OVERVIEW OF SAMPLE AND QUESTIONNAIRE Our main sample was drawn from the total membership list of the Canadian Research Institute for the Advancement of Women (CRIAW). As its name suggests, CRIAW is a national organization devoted to research which will enhance the quality of women's lives. It was FEMINISM AND THE QUALITY OF LIFE 451 founded in 1975 with that goal in mind. CRIAW was chosen as the initial sample because of its large individual membership and because of its goals and aims. Although the Canadian National Action Committee on the Status of Women (NAC) has a larger constituency, it is primarily a lobby rather than a research organization. Also, NAC has only group memberships. Some of these groups are not clearly committed to women's status per se (e.g., a number of traditional faiths have NAC membership). Because we were interested in obtaining a list of items that would be fairly representative of what concerned, wellinformed and thoughtful Canadian feminists believe about feminism rather than merely a list of items indicating what randomly selected Canadian women might happen to believe about feminism, the CRIAW group seemed to be perfect for our purposes. In the fall of 1985 we designed a 14 page questionnaire and pretested it on 15 members of CRIAW-Nova Scotia. Responses to the pretest resulted in a revised questionnaire which was sent to 709 national CRIAW members. Address labels were provided by the CRIAW national office. Duplicate sets of labels were identically numbered from 1 to 709. In order to ensure anonymity, a research assistant who was not familiar with the names of individual members of CRIAW was employed to check-off names as questionnaires were returned and to send out follow-up letters to those individuals who did not return the questionnaire. The distribution of responses to these mailings were as follows: N % 450 8 ( 63.5) ( 1.1) 1 203 ( 0.1) ( 28.6) Completed questionnaires (by CRIA W members) Questionnaires duplicated by CRIAW members for friends or coworkers Letter returned by post-office Duplicate labels Acknowledged letter, refused to respond Returned questionnaire unanswered because respondents were unilingual Francophone Person died, survey returned by executrix Letter not returned, no response 717 (101.0) Total 19 5 23 8 ( ( ( ( 2.7) 0.7) 3.2) 1.1) 452 DEBORAH C. POFF AND ALEX C. MICHALOS As you will note the total number of questionnaires accounted for is 717. This is due to the fact that some people chose to duplicate the questionnaire and distribute it to co-workers and friends. This accounts for 8 additional returned questionnaires and was only one of the unusual patterns of response among the CRIAW membership. We left these 8 questionnaires in the sample because all of the respondents identified themselves as feminists. The total number of completed questionnaires which we received was 458. Two males completed the questionnaire and these were deleted from the sample leaving us with 456 cases. In addition, we deleted those surveys which had greater than 10 percent missing values. This left us with 439 workable cases, which is 61 percent of the 717 questionnaires; or 431 workable CRIAW cases, which is 61 percent of the original 709 questionnaires. Responses to the demographic items in our questionnaire indicated that 422 respondents were aged from 21 to 77, with a mean of 42 and a median of 40. Fifty-nine percent were married or living with a partner; 70 percent were employed full-time; 74 percent had completed at least a Master's degree and 39 percent a Ph.D.; 92 percent were Canadian citizens, and 65 percent reported family (or equivalent) income of at least $40 000. (See Table I for details). The substantive items in our questionnaire were divided into three parts. Part I had 110 items with three possible responses, namely, "Basically agree", "Basically disagree", and an off-scale option saying "Undecided". Respondents we~ asked to check the appropriate answer, and we scored agreement with 2 points and disagreement with 1. Ninety-seven items were selected for their face validity or apparent relevance to a contemporary feminist social, political and economic agenda. The agenda was a product of our own reflections and assessments of current literature and activities of self-avowed feminists. Most of the items were invented, but some were borrowed from earlier studies. The other 13 items in this part of the questionnaire included four borrowed from the self-esteem index of Rosenberg (1965) and 9 from the political conservatism scale of McClosky (1958). Part II had eight items in the same basic format as the earlier items, but with an additional off-scale item saying "Doesn't apply to me" for respondents who were unemployed or had no spouse or living partner. 453 FEMINISM AND THE QUALITY OF LIFE TABLE I Sample composition Age N % Nearest family Income N % 21-30 31-40 41-50 51-60 61-70 70-77 Total 46 170 126 50 27 3 422 10.9 40.3 29.9 11.8 6.4 0.7 100.0 Marital status N % Up to $9999 $10000-14999 15000-19999 20000-24999 25000-29999 30000-34999 35000-39999 40000-49999 50 000 or above Total 12 14 8 22 25 26 41 74 203 425 2.8 3.3 1.9 5.2 5.9 6.1 9.6 17.4 47.8 100.0 Single Married/Partner Widowed Separated Divorced Total 98 252 13 16 51 430 22.8 58.6 3.0 3.7 11.9 100.0 Country Status N % Canadian citizen Landed immigrant Visa Total 402 35 0 437 92.0 8.0 0 100.0 Employment status N % Unemployed Employed Full-Time Employed Part-Time Retired Student Other Total 13 304 45 16 34 27 439 3.0 69.2 10.3 3.6 7.7 6.2 100.0 Highest level of education N % Grade 12 or 13 2 yr. college degree 3 yr. college degree 4 yr. college degree Master's degree Professional degree Doctorate Other Total 10 8 21 58 121 34 170 17 439 2.3 1.8 4.8 13.2 27.6 7.7 38.7 3.9 100.0 454 DEBORAH c. POFF AND ALEX C. MICHALOS Part III contained the 60 items from the Bern Sex-Role Inventory (Bern, 1976) with three response categories, namely, "Never or usually not true", "Occasionally true" and "Always or usually true". These were scored 1 point for "never", 2 for "occasionally" and 3 for "always". 4. RESPONDENTS AND NON-RESPONDENTS One thing that is patently clear is that the majority of CRIAW members care a great deal about issues related to women's status. Many of the respondents included personal letters that spoke positively about the project. One such respondent ended her questionnaire with the comment "Best of luck with what I'm sure will be a major contribution to Women's Studies!" Others were suspicious and, in some cases, hostile to any attempt to quantify feminist beliefs. Some respondents stated their qualifications in objecting to the project. One such respondent said that she had a Ph.D. in Sociology and consequently found the project naive. This was especially interesting since another Sociologist cited her training and expertise in stating how good she thought the questionnaire was. One respondent circled the item, "I believe in equality of opportunity" beside which she commented, "I wrote an MA. thesis on this topic and it is too complex to address in a single statement". Some worried about the possible abuses of such a study but filled in the questionnaire as an act of faith in the integrity of the researchers. As one respondent wrote, "I don't have faith in your project but 1 do trust you". Many respondents wrote running commentaries along with their answers. Many included personal notes about their lives. One respondent stated that she thought it was important for us to know that although her income looked good, she worked at one full-time job and two part-time jobs to earn enough money to support her two children. Some respondents chose to tell us that the Bern scale was not a good one, assuming that we were unfamiliar with the criticisms of the scale. As one respondent put it, "Shame on you" for including the BRSI. Some tried to second-guess our motives for including the BEM scale, arguing that it was dangerous to try to prove that feminists were androgynous. Our non-respondents can be divided into three categories, those who FEMINISM AND THE QUALITY OF LIFE 455 did not respond at all, those who returned the questionnaire because it was only in English, and those who chose to tell us why they disagreed with the questionnaire and consequently would not fill it in. Of the non-respondents who were unilingual French-speaking, the majority expressed regret that they could not be a part of the project, while one was hostile and remarked that we should realize that we lived in a bilingual country. Our very limited budget precluded translation of the questionnaire, but we regret that this did result in some respondents being unable to participate. Of the 23 non-respondents who corresponded with us, 2 wrote that they were unable to participate in the survey because of their positions with the government. Both were in positions related to the status of women and said that they could not answer questions which would reflect their personal values. Of the other 21 respondents, most believed that the survey was either not adequately designed or that the project was misguided. One such non-respondent stated, I do not believe that the question of the characteristics of feminists arises from a rational stance (i.e., I believe it is an emotional response to, amongst other things, change) and hence, I do not believe that a reasoned, statistical, even, description of feminists is a response that will make a difference. Therefore, I have chosen NOT to fill out the questionnaire. Another non-respondent stated, My main concern is that I am not sure that one can obtain any meaningful 'standardization of measurement' about feminists and feminism. The statements in Part I and Part IT all deal with issues of interest but are in and of themselves meaningless. It is not possible, in my opinion, to take a statement out of a broader context and retain any meaning. Part III of the questionnaire in particular concerns me as I don't feel that there is any point in trying to ascribe any personality or behavioural characteristics to feminists as a whole. In ·fact the danger in this is that the focus on feminists will be on our behavioural characteristics and not on the validity of the arguments we make. While I do see a role for quantitative analysis in social sciences, it can often obscure rather than clarify issues. A more detailed qualitative study of opinions on particular issues would be more useful here, especially one which leaves room for a variety of opinions recognizing that continuing dialogue on issues of interest is what makes any movement vital. The long quotations above seem characteristic of two types of objections that were frequently cited. The first (if we understand it correctly) is a challenge on psychological rather than theoretical grounds. The argument is that those who attempt to classify feminist beliefs do so out 456 DEBORAH C. POFF AND ALEX C. MICHALOS of an irrational fear of feminists or what they stand for. As feminist researchers, we can only say that this was not our motivation for this study. The second challenge we believe to be a more serious and more frequent criticism of attempts to carry out quantitative research on women's issues. The specific criticism is that because we have listed women's issues in a series of discrete statements to which we ask feminists to respond, our statements are "meaningless" and without context. The charge of decontextualization is, in this case, questionbegging at best. That the author's three paragraph letter is contextualized while 110 statements which relate to women's status are decontextualized needs to be proven. As feminists, we are well aware of the danger of decontextualized "male-stream knowledge". However, to suggest that a questionnaire consisting of a number of statements relating to the status of women, variously drawn from earlier studies by feminists or constructed for this occasion by feminists, funded by a feminist institute and administered to one feminist organization is somehow lacking a clear context or meaning is completely without foundation at best and dangerous to feminist research at worst. 5. MISSING VALUES There were 88 items in the original set from which a subset was to be selected to form a feminism scale. In order to ensure that each item we were working with was significant to most respondents in the sense that they had actually made a considered judgment about it, we began by deleting all items that had over 10% missing values. This had the fairly devastating result of reducing our working set to 29. After some preliminary work with this set, we loosened up our criterion of admission to 15% and increased our working set to 39. Roughly speaking, this meant that if more than 65 respondents did not basically agree or disagree with an item, the item was deleted. (For the record, a criterion of 20% would have added another 16 items, and 25% would have 12 beyond that, i.e., 67 (76%) of the original 88 items were answered by about 75% of our respondents.) Besides losing items as a result of our conservative criterion of item admission, we also lost cases (respondents) as a result of our selection 457 FEMINISM AND THE QUALITY OF LIFE of the SPSSx option of deleting cases with missing values in listwise fashion. This is also a conservative strategy, since the listwise deletion option deletes a case from an analysis if any of its relevant variables is missing. Accordingly, varying sample sizes (Ns) will be indicated in the remaining sections of the paper. 6. ASSOCIATIONS AMONG TYPES OF FEMINISM Table II gives a cross-tabulation of six types of feminism, namely, feminists, liberal feminists, socialist feminists, Marxist feminists, radical feminists and lesbian feminists. The purpose of this table is to show the TABLE II Cross tabulation of types of feminism by numbers and percentages a GROUP 1 1. Feminist N Percent 2. Liberal Feminist N Percent 3. Socialist Feminist N Percent 4. Marxist Feminist N Percent 5. Radical Feminist N Percent 6. Lesbian Feminist N Percent GROUP 2 1 2 3 4 5 6 417 422 99 117 314 37 188 313 60 35 326 11 104 337 31 41 369 11 185 325 57 69 276 25 10 284 4 26 283 9 8 300 3 192 323 59 33 273 12 55 274 20 20 297 7 36 341 11 16 308 5 9 327 3 104 352 30 25 332 8 41 384 11 Percentages indicate ratios of numbers of women identifying themselves as members of one or two groups to the total numbers responding to the question, i.e., Group 1 X 1/N, Group 1 X 2/ N, etc. Decimal points are omitted. a 458 DEBORAH c. POFF AND ALEX C. MICHALOS number and percent of self-avowed feminist respondents of various types, and to indicate the sizes of pairs of these types. The entries in the diagonal line from left to right give the number and percent of each type of feminist taken by itself. Thus, we had 417 of 422 or 99% of our respondents regarding themselves as feminists, and so on. Reading across the top row, one finds that 314 respondents answered the feminist and liberal feminist items, and 117 (37%) of these people regarded themselves as not only feminists but also liberal feminists; 188 of 313 (60% ) were feminists and socialist feminists, etc. One can of course speculate about the non-responses to these items, and inflate or deflate figures to fit one's speculations. However, taking the figures at face value, there is substantial evidence that this overwhelmingly feminist group (99%) is decidedly not Marxist (11 %) and not lesbian (11 %), and not even very radical (30%). Table III gives a cross-tabulation of measured levels of association among the six types of feminism. As it turned out, whether one measured these associations using Pearson's point-biserial r, phi or eta, the results were virtually always identical and relatively unremarkable. Ten of the 15 correlations are significant at the 1% level or better. A good sense of the limitations of measures of association given extremely TABLE III Associations among main dependent variables (etas e) Independent Dependent Variables 1 1. 2. 3. 4. 5. 6. a b c Feminist Liberal F. Socialist F. Marxist F. Radical F. Lesbian F. 2 3 4 5 6 04 16 c 19 b 04 16 c 33 a 08 30 8 22 8 22 8 04 -19 8 06 19 8 38 8 P < 0.001 P < 0.005 P < 0.01 Although eta values run from 0 to +1, decimal points are omitted in this table and significant negative relationships are indicated with minus signs. e FEMINISM AND THE QUALITY OF LIFE 459 skewed distributions may be obtained by noticing, for example, that although 60% of those responding to the feminist and socialist feminist items claimed to be both, the measured correlation between these items is only 0.16 (p < 0.01). If one makes the-standard assumption that the squared correlation between two variables is a measure of the percent of variance in one that is explained by the other (i.e., the standard interpretation of the coefficient of determination), the apparently robust relation between the two items in question looks even smaller. On this reckoning, 3% of the variance in socialist feminist scores can be explained by feminist scores. Given the egalitarian bases of feminism and socialism, one would think that the 60% figure we began with is more indicative of the relation between these two items than the 3% figure. Presumably it is the extreme skew and small variance in the items that is creating the anomaly. (There is a good discussion of decision criteria that might be used to select appropriate coefficients of determination in certain situations in Ozer, 1985. Unfortunately, the criteria do not seem to provide much guidance in the present case.) 7. STRUCTURE OF THE PMF SCALE While single items are adequate for identifying self-avowed feminists and have relatively transparent face validity, they are practically worthless from the point of view of content-validity and construct-validity. They reveal nothing about the contemporary feminist agenda or about how adherence to that agenda might be related to a variety of other aspects of one's life. To investigate these issues, we developed the PMF scale. Our strategy was to build up the scale by obtaining Cronbach alpha coefficients for subsets of items that were conceptually related. The idea was suggested by Nunnally (1967), and made operationally efficient by the splendid display of information available with the Reliability procedure of SPSsx. Essentially the alpha coefficient is a measure of the internal consistency of a set of items. Its values range between 0 and 1, reaching the former when no item in the set is correlated with any other and reaching the latter when every item in the set is perfectly correlated with every other. Generally speaking, one would like to have a scale consisting of a set of items with an alpha of 0.8 or above. How- 460 DEBORAH C. POFF AND ALEX C. MICHALOS ever, lower values may be tolerated in the interest of obtaining a relatively high level of content-validity. For example, the following three items of the PMF scale form a nice subset with an alpha of 0.9: Knowledge is male-biased; Morality is male-biased; Language is malebiased. Although the alpha of the PMF scale is considerably lower than this at 0.61, one would have a seriously inadequate measure and a misleading view of feminism if one adopted the nice triad as a scale of feminism. (An excellent discussion of the nature and uses of alpha may be found in Zeller and Carmines, 1980.) Table IV lists the 23 items of the PMF scale, with the numbers and percents of respondents basically agreeing or disagreeing with each item. Sixteen or 70% of the items had majorities of 98% or more either agreeing or disagreeing. Even the item with the least amount of skew (knowledge is male-biased) had 82% agreement. Obviously such skewed distributions can create problems with statistical procedures that are typically designed assuming normal distributions. One notable example is the deletion of items 22 and 23 from scale calculations because both items had zero variance. Since alpha coefficients are based on the covariation of variables, if the latter have no variance then they can not have any impact on the coefficient. So they are routinely deleted from scale calculations. However, if virtually all feminists agree that sexual harassment includes threats of demotion or promises of promotion in exchange for sexual favours (item 22) and that women and men should be equally responsible for all domestic labour (item 23), then any reasonable index of feminism should give respondents the opportunity to affirm these propositions. Thus, we have included them in the PMF scale, although they played no role in the SPSSx statistical calculations. Broadly speaking, the 23 items of the PMF scale cover the areas of workplace equality (items 1 and 2; a ... 0.61), parenting and domestic labour equality (5, 23, a ... 0.44), abortion rights (7, 8; a - -0.60), wife abuse (10, 11; a - 0.69), male-biased institutions (14, 15, 16; a -= 0.90), sexual harassment (18, 19, 20, 21, 22; a -= 0.78), need for unity (3), control of one's own body (6), social security rights (4), midwifery access (9), rape responsibility (12), day care access (13) and justice for all (17). In still broader terms, the 23 items may be said to cover the areas of socio-economic justice and the nature of the oppres- 461 FEMINISM AND THE QUALITY OF LIFE TABLE IV Numbers and percent of respondents basically agreeing or disagreeing with items in the PMF scale Item 1. I believe in equality of opportunity in the workplace. 2. I believe in equal pay for work of equal value. 3. Women need to unite and work together to achieve equal political and social rights in this country. 4. Quebec Pension Plan or Canada Pension Plan benefits should be available to all women at retirement age. 5. Both male and female parents should be equally responsible for the care of young children. 6. A woman should have total control over her body, including reproductive control. 7. Abortion is never justifiable. 8. A woman always has the right to have an abortion. 9. Midwifery should be legal and accessible to all women. 10. Wife abuse is never justifiable. 11. Men are innately violent and consequently beat their wives. 12. Women should be held partially responsible for rape. 13. There should be only private day care. 14. Knowledge is male-biased. 15. Morality is male-biased. 16. Language is male-biased. 17. I believe in justice for all. 18. Sexual harassment includes having to listen to sexist jokes at work. 19. Sexual harassment includes unwanted remarks about your appearance. 20. Sexual harassment includes pornography on a male colleague's desk at work. 21. Sexual harassment includes unnecessary touching by male workers or employers. 22. Sexual harassment includes threats of demotion or promises of promotion in exchanges for sexual favours.* 23. Women and men should be equally responsible for all domestic labour.* Agree (%) N N Disagree (%) 432 425 (98 .9) (99.1) 5 4 (1.1) (0.9) 423 (99.5) 2 (0.5) 402 (96.9) 13 (3.1 ) 428 (98.8) 5 (1.2) 412 7 (98.1) (1.7) 8 417 (1.9) (98.3) 339 (89.7) 39 (10.3) 404 422 (98.5) (97 .9) 6 9 (1.5) (2.1 ) 9 (2.3) 387 7 5 324 324 375 422 (1.7) (1.2) (81.8 (84.2) (90.4) (98 .8) 409 412 376 (93.3) 27 (6.7) 365 (92 .9) 28 (7.1) 350 (87 .9) 48 (12.1) 428 (99.5) 2 (0.5) 435 (100.0) 0 (0.0) 426 (99.5) 2 (0.5) (97.7) (98.3) (98.8) 72 (18.2) 61 (15.8) 40 (9.6) 5 (1.2) * Items with zero variance were not included in scale calculations but should be included in applications of the scale. 462 DEBORAH C. POFF AND ALEX C. MICHALOS sion of women. In brief, the scale contains a negative assessment indicating female oppression and a positive vision of a world without such oppression; in a word, it contains a contemporary feminist agenda. Table V gives the means, standard deviations and correlations of each of the 21 workable scale items with the total scale score. As mentioned above, the alpha coefficient of 0.61 is not as big as one would prefer. Similarly, the mean inter-item correlation of 0.06 is hardly robust. Maintaining such a low inter-item correlation, one would need around 100 items in order to increase the alpha value above 0.80. So there is clearly room for further development regarding both the number and kinds of items, and the PMF scale should be regarded as a beginning rather than an end. While it is likely that a living, growing and dynamic socio-political movement like feminism will virtually always contain discordant beliefs, attitudes, assumptions and visions, we still believe that further efforts to identify a coherent set of shared ideas will be rewarded with higher alpha and inter-item correlation levels. If anyone responded to the 21 items exactly as the majority of people responded, her score would have been 38. Although the scale mean was 37, Table VI shows that only 2% of respondents scored 37. Sixty-eight percent scored 42. (Because of automatic rounding off of scores, values up to 44 were possible depending on how variables with zero variance were treated.) 8. VALIDITY OF THE PMF SCALE "Validation", Nunnally (1967) has wisely written, "is an unending process." Here we are interested in initiating the validation of the PMF scale in four senses of this term (following Nunnally, and Zeller and Carmines, 1980). A measure (scale, or index) has criterion-related validity insofar as it has a significant correlation with some other relevant measure or criterion. For example, the written driver's test one must pass in order to take a road test is criterion-valid exactly insofar as the former is correlated with the latter. As this example indicates, nothing about the validated test (e.g., sitting at a desk answering a multiple choice examination) has to look like anything in the criterion test (driving a car). The only requirement is that results of both tests or measures are correlated. 463 FEMINISM AND THE QUALITY OF LIFE TABLE V Means, standard deviations and item-total score correlations for items of the PMF scale (N = 200) Item 1. I believe in equality of opportunity in the workplace. 2. I believe in equal pay for work of equal value. 3. Women need to unite and work together to achieve equal political and social rights in this country. 4. Quebec Pension Plan or Canada Pension Plan benefits should be available to all women at retirement age. 5. Both male and female parents should be equally responsible for the care of young children. 6. A woman should have total control over her body, including reproductive control. 7. Abortion is never justifiable. 8. A woman always has the right to have an abortion. 9. Midwifery should be legal and accessible to all women. 10. Wife abuse is never justifiable. II. Men are innately violent and consequently beat their wives. 12. Women should be held partially responsible for rape. 13. There should be only private day care. 14. Knowledge is male-biased. 15. Morality is male-biased. 16. Language is male-biased. 17. I believe in justice for all. 18. Sexual harassment includes having to listen to sexist jokes at work. 19. Sexual harassment includes unwanted remarks about your appearance. 20. Sexual harassment includes pornography on a male colleague's desk at work. 21. Sexual harassment includes unnecessary touching by male workers or employers. 22. Sexual harassment includes threats of demotion or promises of promotion in exchanges for sexual favours.* 23. Women and men should be equally responsible for all domestic labour.* Means St. Dev. Corr. 1.98 1.99 0.144 0.070 0.18 0.07 1.99 0.070 0.29 1.98 0.140 0.15 1.99 0.122 0.26 1.98 1.01 1.91 0.140 0.070 0.294 0.15 -0.02 0.16 1.99 1.98 0.122 0.140 -0.06 0.15 1.03 0.157 0.04 1.02 1.02 1.87 1.88 1.93 1.98 0.122 0.122 0.337 0.326 0.256 0.140 -0.24 -0.06 0.45 0.48 0.36 0.15 1.95 0.219 0.33 1.96 0.208 0.38 1.92 0.280 0.39 1.99 0.071 0.24 2.00 0.000 2.00 0.000 * Items with zero variance were not included in scale calculations but should be included in applications of the scale. Scale mean = 37.32, St. Dev. = 1.33; Mean of item means = 1.78; Mean of item variances = 0.04; Mean of inter-item correlations = 0.06; Cronbach Alpha Coefficient = 0.61. 464 DEBORAH c. POFF AND ALEX C. MICHALOS TABLE VI . Distribution of respondent PMF scale scores by score value, frequency number and percent Score value Frequency number Frequency percent Cumulative percent 44 43 42 41 40 39 38 37 36 35 Total: 1 2 136 23 13 18 1 4 1 1 200 0.5 1.0 68.0 11.5 6.5 9.0 0.5 2.0 0.5 0.5 100.0 100.0 99.5 98.5 30.5 19.0 12.5 3.5 3.0 1.0 0.5 A measure has content-validity insofar as it adequately or completely refers to the relevant content of some area or domain to be measured. We have already mentioned several examples of measures of feminism that seem to have not only limited but seriously misleading contentvalidity. Zeller and Carmines (1980) claimed that "The major problem associated with this type of validity is that there are no agreed-upon criteria for establishing whether, in fact, a measure has attained content validity." And Nunnally (1967) wrote that In spite of efforts on the part of some to settle every issue about psychological measure- ment by a flight into statistics, content validity is mainly settled in other ways. Although helpful hints are obtained from analysis of empirical findings, content validity mainly rests upon an appeal to the propriety of content and the way that it is presented. Hence, we have been inviting readers to reflect upon the agenda articulated in the content of the items of the PMF scale, and we will have nothing else to add in support of the content-validity of this scale. A measure has construct-validity insofar as it has significant correlations with some other theoretically relevant measures. Since measures like the PMF scale are literally constructed in the interest of exploring, confirming or disconfirming more or less developed theories about certain features of the world, some assessment of the validity of such measures must involve examining their theoretical relationships. For 465 FEMINISM AND THE QUALITY OF LIFE example, we began this research with the theoretical hypotheses mentioned in the first section, i.e., that feminists are generally good people, equally concerned with their careers and families if they have both, happy, and so on. These hypotheses are tested below. A measure has discriminant-validity insofar as it may be used to discriminate individuals in a random sample on the basis of their having or lacking the property purportedly being measured. Hence, in order to show that the PMF scale has discriminant-validity, it was applied to a new sample as described below. Results of testing the criterion-related and construct-validity of the PMF scale are summarized in Table VII. The most salient criterion available is evidently the single item asking respondents if they consider themselves to be feminists. Table VII indicates that the zero-order correlation of this criterion with the PMF scale is 0.26 (P < 0.001). This confirms hypothesis Hla. Although this figure looks low, it may be increased by correcting it for attenuation resulting from the relative unreliability (imperfection) of the correlated measures. The procedure is worth illustrating in spite of its inherent limitations (indicated below). Because, as Carmines and Zeller (1979) succinctly say, "The square root of the reliability of a measure provides an upper bound for its correlation with any other measure" (i.e., r(xy) ~ Jr(xx)), one can make an estimate of the reliability of a measure on the basis of its correlation with some other measure. For the case at hand, because the TABLE VII Pearson product moment correlation of PMF scale with selected measures Item or Index r Sig. N Feminism Radical Feminism Humanism Good Person Index Independent Happy Truthful Compassionate Friendly Competitive CareeriFamily Importance 0.26 0.23 -0.15 0.20 0.15 0.19 0.15 0.20 0.18 -0.15 0.22 0.001 0.005 0.05 0.01 0.05 0.01 0.05 0.005 0.01 0.05 0.01 198 165 177 189 195 195 193 194 194 194 143 466 DEBORAH c. POFF AND ALEX C. MICHALOS feminist criterion had a correlation of 0.63 with item # 3 on the PMF scale, we know that its reliability is at least 0.63 2 or 0.40. Since the alpha coefficient is a standard measure of reliability, we may apply the following formula to correct our correlation for attenuation: r(xy)* = _ _r--,,--(x--=--YL-)_ Jr(xx)r(yy) (Zeller and Carmines, 1980) In this formula r( xy)* represents the corrected correlation coefficient, r(xy) represents the uncorrected coefficient, r(xx) and r(w) represent the reliability coefficients of x and y, respectively. Thus, for the case at hand, we have r(xy)* = _ _0_.2_6_ _ = ~(0.61)4 0.26 0.49 = 0.53 ' which is a bit better than the original 0.26. The trouble with this approach to estimating the reliability of a measure is that it tends to capitalize on ignorance and can lead to absurd results. For example, the PMF scale has a correlation of 0.23 with radical feminism, but because the latter's highest correlation with another variable so far as we know is 0.41 (with separatist feminism), correcting for attenuation yields a correlation coefficient of 0.72. Unfortunately, if the highest known correlation of radical feminism with another variable had been 0.20, the corrected correlation coefficient would have been 1.43, which is absurd. In the light of such anomalies, it is best to urge caution with the strategy suggested above. For our purposes, it is enough to know that the PMF scale has the hypothesized statistically significant correlation with the most salient criterion available and with another species of feminism. We had hoped to be able to identify all types of feminism with some subset of all our items, if not from a subset of the 23 items in the PMF scale. But this proved to be impossible. Results of testing the construct-validity of the PMF scale were fairly encouraging. As indicated in Table VII and hypothesized, there were statistically significant positive correlations between being a feminist as measured by the PMF scale and describing oneself as happy (confirming Hlb), independent (HIc), truthful (HId), compassionate (HIe) and 467 FEMINISM AND THE QUALITY OF LIFE friendly (Hlf). The latter four attributes were added to seven others to form an index of what we called simply "a good person". Table VIII lists the total set of attributes of our good person index and indicates its alpha coefficient of 0.74. As hypothesized in Hlg, there was a significant positive association between the PMF scale and good person index scores (Table VII). Similarly, we confirmed the hypothesis (Hlh) that PMF scale scores would be positively related to scores indicating that one's family and career were equally important. PMF scale scores were found to be negatively associated with describing oneself as competitive (H2). TABLE VIII Means, standard deviations and item-total score correlations for good person index· Item Mean St. Dev. Corr. Self-reliant Helpful Defends own beliefs Cheerful Independent Reliable Truthful Compassionate Sincere Likable Friendly 2.91 2.79 2.90 2.72 2.93 2.95 2.93 2.75 2.92 2.65 2.82 0.286 0.411 0.308 0.466 0.290 0.229 0.263 0.457 0.280 0.508 0.400 0.25 0.42 0.15 0.50 0.33 0.25 0.31 0.49 0.50 0.52 0.53 • Index mean = 31.26, St. Dev. = 2.13; Mean of Items = 2.84; Mean of Item Variances = 0.13; Mean of Inter-Item Correlations = 0.20; Cronbach Alpha Coefficient = 0.74. (N = 402) Our attempt to construct a scale of self-esteem using the four positive items of Rosenberg's scale failed because three of the four items had no variance at all. It is instructive to consider the four items and the distribution of responses to them. These are given in Table IX. Considering the numbers and percent distributions in this table, one must conclude that the CRIA W women had extremely high self-esteem. In fact, comparing these figures to those regarding their commitment to feminism, it is fair to say that these women were even more certain of their own self-esteem than they were of their feminism. Given the 468 DEBORAH C. POFF AND ALEX C. MICHALOS TABLE IX Numbers and percent of respondents basically agreeing or disagreeing with positive self-esteem items Item 1. I feel that I'm a person of worth, at least on an equal basis with others. 2. I feel that I have a number of good qualities. 3. I am able to do things as well as most people. 4. I take a positive attitude toward myself. N Agree (%) Disagree N (%) 0 (100.0) o o o (98.8) 5 (1.2) 434 (100.0) 436 (100.0) 433 420 0 0 number of studies indicating that women in general tend to have relatively low self-esteem, the fact that practically 100% of this feminist group reported high self-esteem is both remarkable and encouraging. Our attempt to construct a scale of political conservatism using the McClosky items failed because most of the items had over 15% missing values. However, it has already been shown that a majority of the group leaned toward the left side of the political spectrum. There was no significant association between being a feminist as indicated by the PMF scale and describing oneself as yielding, gullible, childlike, conventional, aggressive, conscientious, assertive, loyal, understanding, warm, willing to take a stand, individualistic, having leadership abilities or loving children (H3). Our tests of these several hypotheses were straightforward and similar to the preceding. We simply posited a zero-order correlation and made correlational analyses with two-tailed significance tests. While we usually had a preference for finding a particular kind of association, either positive or negative, it seemed most efficient to make the weakest predictions possible. If these failed, then questions concerning the direction of relations would be pointless, and if they succeeded, we would have answers to the directionality question immediately from the positive or negative correlation coefficients. To test the discriminant-validity of the PMF scale, we administered a new questionnaire to a convenience sample of 440 undergraduate FEMINISM AND THE QUALITY OF LIFE 469 women at the University of Guelph. The women were students in 9 different classes, most of which were second or third year courses. While the sample cannot be regarded as representative of Canadian women in general, it is very probably a fair sample of undergraduate women at Guelph and probably at most other Canadian universities. The questionnaire contained the 23 items of the PMF scale and five other items asking if respondents considered themselves to be humanists, liberals, feminists, conservatives or socialists. For all items the response options were "basically agree" or ''basically disagree". There were 413 useable returns, and 159 (39%) of the respondents considered themselves to be feminists. Thus, insofar as the PMF Scale could successfully discriminate those individuals from the non-feminists, its discriminant-validity would be confirmed. Since 97%· of the CRIAW sample scored from 39 to 44 points on the PMF scale, we used scores in this range as clear indicators that respondents had feminist beliefs and attitudes. Then, we constructed a simple 2 X 2 contingency table cross-tabulating self-avowed feminism with feminism as measured on the PMF scale. A Chi-square test showed that the PMF scale scores could be used to appropriately sort the undergraduate sample (Chi-square ~ 23.97, 1 df, P < 0.0001). Hence, the discriminant-validity of the PMF scale was confirmed. Inspection of the various percentages of undergraduates and CRIAW members agreeing or disagreeing with items in our scale revealed that the biggest discrepancies were related to the three items concerning the male-biased nature of knowledge, morality and language. So, if one were only interested in a scale that would successfully sort feminists from non-feminists (i.e., only in a scale with discriminant-validity), a scale based on this triad would be a fine candidate. Its Cronbach alpha, you may recall, is 0.90. 9. CONCLUDING REMARKS Although the PMF scale has been shown to have at least acceptable levels of reliability, criterion-related, content, construct and discriminant-validity, there is still much to be done. We intend to increase its content-validity by experimenting with additional items that provide a more thorough articulation of the contemporary feminist agenda. While 470 DEBORAH C. POFF AND ALEX C. MICHALOS the contemporary Canadian feminist agenda is not identical to, say, its nineteenth century counterpart and its counterparts in other nations today, we believe it is similar to both. We think it should be possible to construct an index that has a central core of items indicating a feminist vision of social, economic and political justice that would be shared by virtually all feminists everywhere and always. We intend to test this belief empirically by experimenting with the PMF scale and its offspring in diverse cultures over the next few years. There were several items in our questionnaire that we regarded as good candidates for a contemporary Canadian feminist agenda, but which were not similarly regarded by the CRIAW respondents. This is not the place to review arguments in support of or opposed to such items, but it may be worthwhile to simply mention some of these items in order to provoke further discussion about them. For each of the following items, respondents' answers were distributed relatively evenly among the three options "Basically agree", "Basically disagree" and "Undecided". 1. Quota systems should be legalized and enforced until women have equal representation in all occupations. 2. All censorship is dangerous. 3. Prostitution should be legal and should be treated as all other paid employment with the same rights and restrictions. 4. The breakdown of the nuclear family would be a real loss to society. 5. It is impossible to be a feminist and believe in the doctrines of traditional religious faiths. Clearly, these are complex issues and reasonable people could have good reasons for being for one side or another, or for being completely baffled and undecided. Whatever the reasons were for the random-like distribution of responses to these items, it is important to notice that there are some controversial issues yet to be resolved within the feminist movement and it is important to be able to identify them. As we suggested in the introduction to this paper, the clarification of our shared feminist vision is absolutely essential for the continued development of progressive scientific research, social, political and economic programs. FEMINISM AND THE QUALITY OF LIFE 471 ACKNOWLEDGEMENTS We would like to express our thanks to all the members of CRIAW who so generously took the time and effort to answer our survey, to AI Dyer for his help with the computer work in this project, Christina Schraefel who spent many long hours entering the data, Marirose McCabe who typed the various drafts of the paper, and to Karl Schuessler for helpful suggestions. Earlier versions of the paper were presented at the annual meeting of the Canadian Sociology and Anthropology Association, Winnipeg, June 1986, and to The International Sociological Association meeting, New Delhi, August 1986. REFERENCES Acker, J. et al.: 1974, 'A feminism scale: A report on its construction', UnpUblished manuscript available from the Center for the Sociological Study of Women, University of Oregon. Alspach, S.: 1982, 'Women's sex role attitudes and life satisfaction', Sociological Focus 15,279-287. Aslin, A. L.: 1977, 'Feminist and community mental health center psychotherapists' expectations of mental health for women', Sex Roles 3, 537-544. Beere, C. A. et al.: 1984, 'The sex-role equalitarianism scale: A measure of attitudes toward equality between the sexes', Sex Roles 10,563-576. Bem, S. L.: 1976, 'Probing the promise of androgyny', in Beyond Sex-Role Stereotypes: Readings Toward a Psychology of Androgyny, ed. by A. G. Kaplan and J. P. Bean (Little, Brown and Co., Boston), pp. 48-62. Berryman-Fink, C. and K. S. Verderber: 1985, 'Attributions of the term feminist: A factor analytic development of a measuring instrument', Psychology of Women Quarterly 9, 51-64. Cannines, E. G. and R A. Zeller: 1979, Reliability and Validity Assessment (Sage Publications, Beverly Hills). Dempewolf!, J. A.: 1974, 'Development and validation of a feminism scale', Psychological Reports 34,651-657. Foster, M. A., B. Strudler-Wallston and M. Berger: 1980, 'Feminist orientation and jobseeking behavior among dual career couples', Sex Roles 6, 59-65. Kirkpatrick, c.: 1936, 'The construction of a belief-pattern scale for measuring attitudes toward feminism', Journal of Social Psychology 7, 421-437. McOosky, H.: 1958, 'Conservatism and personality', American Political Science Review 52 27-45. Nielsen, J. McCarl and P. Thoits Doyle: 1975, 'Sex-role stereotypes of feminists and nonfeminists', Sex Roles 1,83-95. Nunnally, J. C.: 1967, Psychometric Theory (McGraw-Hill Book Co., New York). Ozer, D. J.: 1985, 'Correlation and the coefficient of determination', Psychological Bulletin 97, 307-315. Poff, D. c.: Nov. 1985, 'Feminism flies too', Resources for Feminist Research. Singleton, R and J. B. Christiansen: 1977, 'The construct validation of a short form attitudes toward feminism scale', Sociology and Social Research 61, 294-303. Smith, E. R, M. M. Ferrez and F. D. Miller: 1975, 'A short scale of attitudes toward feminism', Representative Research in Social Psychology 6, 51-56. Smith, M. D. and G. D. Self: 1981, 'Feminists and traditionalists: An attitudinal companion', Sex Roles 7,183-188. 472 DEBORAH C. POFF AND ALEX C. MICHALOS Wilson. G. D.: 1982, 'Feminism and Marital dissatisfaction', Personality and Individual Differences 3, 345-347. Zeller, R A. and E. G. Carmines: 1980, Measurement in the Social Sciences: The Link Between Theory and Data (Cambridge University Press, Cambridge). Institute for the Study of Women, Mount Saint Vmcent University, Hali/ax, Nova Scotia, Canada B3M 216.