Appendix A: Methods of Research

Extending the discussion I began in the introduction, this appendix describes elements of my method of research on contemporary women's friendships and marriage: respondents I interviewed, interview design, and problems of inquiry and inference that I encountered with this exploratory method (the interview schedule is reproduced in appendix B).

After open-ended "pre-test" interviews with men and women, I collected references from associates acquainted with women outside the sphere of university influence. My associates knew that friendship was the topic of my study. When I asked them for names, I told them I was looking for "ordinary" women; I stressed that I did not prefer especially sociable women, that I would like "just the first people who come to mind" who fit a set of characteristics—a few combinations of class, race, age, and life-cycle stage. Only two of twenty-three women I contacted declined to be interviewed. The women represented a range of statuses, personalities, and styles of friendship. Yet each seemed very much like others of her kind whom I have known in other places and times. This is no representative sample, of course, but it includes a variable range of women (which I described in the introduction).

In designing the interview, I standardized both questions and format across the interviews (in appendix B). I wanted to make it possible for others to replicate this study. I also wanted strong results, assuming that often-repeated themes would allow me to make more precise and confident theories. Finally, my standardized questions permitted maximum

intrusion with maximum ethical control. I believed that a free-form discussion about the very intimate areas of friendship and marriage could easily range beyond areas where my subjects felt comfortable. Long experience in interviewing people about marriage has taught sociologists that many individuals are uneasy talking about sex and money. No comparable lore points to topics of intimate friendship that could cause discomfort or distress. By considering the ethics of each question as I developed it, I hoped to save myself the self-consciousness and distraction of constant ethical conundrums, and instead pay full attention to the woman I was interviewing.

As it turned out, the questions I developed worked very well from the beginning. Initial interviews prompted me to add and subtract a few questions, but I successfully used the remainder with all the respondents. Occasionally, I rephrased a question to be certain I understood the meaning it had for the woman I was interviewing. The answers convinced me that my questions were consistently interpreted by the variety of women I interviewed. To treat the related problem of internal reliability, I repeated some questions as the interview progressed.

Exploratory methods do not produce generalizable conclusions but, especially in an unexplored area, these methods can probe patterns and meanings that more conclusive methods are likely to miss. My favorite example derives from a criticism I received for "confounding" kinship and friendship by accepting some women's choice of sisters as best friends. My critic's suggestion to separate friendship from other kinds of primary relationships was perfectly appropriate. It alerted me to explain (in discussing close networks in chapter 3) how I discovered that women who call sisters best friends are describing relationships that, in important ways, are more like others' best friendships than they are like their own and others' sibling relationships. Had I begun by isolating friendship from kinship, I would have found that many women of few means and large families do not have best friends. But that is far from true. An exploratory method, which allowed careful probing for the unexpected, suggested that some women are more likely to select sisters as best friends, and that these relationships are properly considered best friendships as well.

My research design did leave problems of inquiry and inference, however. I describe these limitations here, to accomplish a second task as well: to remind the reader that the goal of this research is hypothesis-generation and invite a more conclusive investigative sequel.

I begin with a gaping problem of design: the missing men. It is surely obvious that a motivating interest in this research is gender difference and stratification; yet I interviewed only women. My goal of digging deep into the substance of women's friendships is well served by this focus.

But when the revelations about women prompt comparison with men's friendships, I am forced either to call forth evidence from other research—almost none of which is strictly comparable—or to rely on women's beliefs about men's friendships. I find it provocative to do both, but I remind the reader that nowhere in this work are the subjective voices of men.

Second, there is the risk of geographical fallacy: are Californians an American group? I did not aim at a random or representative sample; I selected respondents with a variety of characteristics that I believed would evoke a range of American women. I sampled heavily in two small cities, one suburb, an unincorporated area much like a small town, and finally, a large city. Although all Californians are reputed to have arrived here from somewhere else, about half of my respondents grew up in or near the city where our interview took place (one-fourth of them lived in the town where they grew up; another fourth, less than an hour's drive away; the number of those who had moved within the past five years was just below the national average of 20 percent).^[1] Also defying the California stereotype, among the twenty ever-married women, the ratio of married to divorced women conformed to the national average. A slightly higher than average number of married women were in their first marriages (in 1977, 80 percent of all married couples were in their first marriages).^[2]

I interviewed eighteen white and three black women, but not Hispanic or Asian women, who would have given this sample a richer California verity. I did, however, interview several women of southern white ancestry, a populous California ethnic minority; most of them were second-generation Californians. California has many of the clerical, service, electronics-assembly, and seasonal-agricultural industries that draw female labor forces. Exaggerating these peculiarities of the California labor market, employed women were vastly overrepresented in my sample. So were women with some community college education, which I believe is more pervasive among middle- and working-class Californians because of the state's large, accessible, and, until recently, free community college system. Rather than being representative, my sample re-fleeted a variety of American women's experiences, emphasizing those—like work and higher education—that women, in general, are increasingly undertaking. Some characteristics of my twenty-one respondents may be summarized as follows:

My sampling method introduced systematic biases. By securing names from contacts who knew respondents, I began with a sample of women who were unlikely to be extremely isolated: somebody I knew, knew them. In general, I did not "snowball" sample (that is, interview people whom previous respondents identified) in order to avoid tapping networks of shared values or friendship cultures. I hoped instead to tap a variety of such values and avoid the demographic similarity that is a well-established characteristic of personal networks.^[3]

Still, I chose to snowball sample three times. Twice, I interviewed best friends, to see if comparing friends' testimonies would prove fruitful. In one case, I interviewed two ardent, relatively new, best friends. In another case I interviewed the sister of one respondent, whom the respondent considered her best friend. The sister did not reciprocate the designation, however. So only one pair in my sample exaggerated the strength of some results by adding a second identical response to some questions. The variation among responses that should have been identical hints at the latitude a researcher must allow in factual self-reports. I compensated for these few parallel accounts when I described my findings in qualitative terms, but I did not adjust numbers when I counted responses. In the third case, I snowball sampled when a respondent who was particularly interested in the project thought of an acquaintance who

fitted a number of characteristics I had been unsuccessfully seeking in a respondent. I saw no effect from this particular snowball.

Finally, I noted an interactive bias: it was much easier to establish a productive rapport with and ultimately "get to know" women who had very close friends. I do not wish to exaggerate this point for, with one possible exception, I was gratified to establish a comfortable trust and rapport with the women I interviewed. Still, women who engaged in very close friendships were practiced at endowing trust and sharing personal experience in a way that enabled me to glean a wealth of understanding in a relatively short period. I think this distinction did not significantly affect my conclusions, yet it is a methodological issue worth considering.

I would certainly expect interactive effects when interviewing on a topic like friendship. However loose its normative structure, friendship itself is widely considered a moral good. How could I interview people on this topic, asking them for all kinds of details about their friendships and their feelings about friends, without encouraging them to describe themselves favorably as people who have done and generally do what is good? I will describe the steps I took to limit this tendency, but I believe that only the addition of an unobtrusive method (some kind of hidden observation) would resolve doubts about this problem.

I did not introduce my project as a study of friendship. In my introductory letter, I stated I was conducting "a study of the everyday lives of Americans. The purpose . . . is to learn how friends, relatives, and family fit into the lives of people in various communities." In this way I was able to secure consent to the kind of interview I would conduct, without labeling my central interest. I began my interview with open-ended ambiguous questions about self, family, and friends and then moved to more pointed and specific questions. Occasionally, these unfocused questions elicited attitudes and feelings that women downplayed in more focused contexts. Such questions often made it easier to identify less "approved" themes in more defensively phrased answers. Finally, I asked a series of questions about how people "in general" conduct or feel about friendships. This left an opening for respondents to project inadmissible feelings they would not wish to claim for themselves.

My assessment of this last source of distortion is that it was not great, although I believe it was a faint and elusive note. For the most part, what could have been trite descriptions or sentiments led to detailed accounts of those experiences that proved them spontaneous and sincere and utterly defied a judgment of artificial or forced production. I frequently tested responses by asking for more details; my confidence in the sincerity of answers also grew every time one expressed an attitude that countered my own beliefs or expectations.

One aspect of my research strategy probably limited my findings. I was especially reluctant to probe areas that made women uneasy. At the outset I decided that before I intruded into areas that triggered unease I would have to know much more about the meaning of women's friendships. This is another way in which these interviews must be considered exploratory. I wanted to detect signs of perilous territory in friendships and carefully map the dynamics of friendships for further considerate investigation. A quick "No" that I interpreted to mean "Caution here)" may have indicated a simple moment of impatience rather than a wish to avoid the subject. Next time I might follow a "No" to a question like, Can you think of a time with your best friend when you felt you'd just "had enough"? with probes like, Never? I preferred to conduct a delicate first inquiry, accepting only that which women willingly offered.

Because it is a first inquiry on the relation between women's friendships and marriage, and because friendship has been studied so little, I would have considered a study of just about any size or method to be exploratory. Small-sample in-depth interviews allowed me to probe meanings in culture and dynamics of friendship; this would have been much more difficult with other methods. The same strategies that limited the generalizability of my conclusions opened avenues to new theory. If I have been successful, it is because my methods enabled me to explore some issues in depth and to organize observations on friendship and marriage into theory that will invite future research and that more conclusive methods will modify. The study of friendship is just beginning. In introducing it to sociological considerations of social structures in personal life, I aimed to make it theoretically fruitful.