1. SELECTING PARTICIPANTS
Either during the contact process or shortly thereafter the researcher takes the crucial step of selecting the people he or she will interview. The purpose of an in-depth interview study is to understand the experience of those who are interviewed, not to predict or to control that experience. (See Van Manen, 1990, p. 22, for further comment on this fundamental characteristic of a phenomenological approach to research.) Because hypotheses are not being tested, the issue is not whether the researcher can generalize the finding of an interview study to a broader population. Instead the researcher’s task is to present the experience of the people he or she interviews in compelling enough detail and in sufficient depth that those who read the study can connect to that experience, learn how it is constituted, and deepen their understanding of the issues it reflects. Because the basic assumptions underlying an interview study are different from those of an experimental study, selecting participants is approached differently.
1.1. “Only Connect”
The United States has more than 200,000 community college faculty. In our study of the work of community college faculty (Seidman, 1985), we could interview only 76 of them. The problem we faced was how to select those 76 participants so that what we learned about their experience would not be easily dismissed as idiosyncratic to them and irrelevant to a larger population. In their influential essay on experimental and quasi-experimental design, Campbell and Stanley (1963) call this the problem of external validity.
A conventional way of defining the issue is to ask whether what is learned from the interview sample can be generalized to the larger population. One step toward assuring generalizability is to select a sample that is representative of the larger population. The dominant approach to representativeness in experimental and quasi-experimental studies has been the random selection of participants. Theoretically, if a large enough sample is selected randomly or through a stratified, randomized approach, the resulting participant pool is not likely to be idiosyncratic.
In interview studies, however, it is not possible to employ random sampling or even a stratified random-sampling approach. Randomness is a statistical concept that depends on a very large number of participants. True randomness would be prohibitive in an in-depth interview study. Furthermore, interview participants must consent to be interviewed, so there is always an element of self-selection in an interview study. Self-selection and randomness are not compatible.
The job of an in-depth interviewer is to go to such depth in the interviews that surface considerations of representativeness and generalizabil- ity are replaced by a compelling evocation of an individual’s experience.
When this experience can be captured in depth, then two possibilities for making connections develop. They are the interview researcher’s alternative to generalizability. (See Lincoln & Guba, 1985, for an extensive discussion of the concept of generalization.) First, the researcher may find connections among the experiences of the individuals he or she interviews. Such links among people whose individual lives are quite different but who are affected by common structural and social forces can help the reader see patterns in that experience. The researcher calls those connections to the readers’ attention for inspection and exploration.
Second, by presenting the stories of participants’ experience, interviewers open up for readers the possibility of connecting their own stories to those presented in the study. In connecting, readers may not learn how to control or predict the experience being studied or their own, but they will understand better their complexities. They will appreciate more the intricate ways in which individual lives interact with social and structural forces and, perhaps, be more understanding and even humble in the face of those intricacies.
1.2. Purposeful Sampling
How best to select participants who will facilitate the ability of others to connect if random selection is not an option? The most commonly agreed upon answer is purposeful sampling. Patton’s (1989) discussion of purposeful-sampling techniques is very thoughtful. He suggests several approaches, including “typical case,” “extreme or deviant case,” “critical case,” “sensitive case,” “convenience” sampling, and “maximum variation” sampling (pp. 100-107).
Maximum variation sampling can refer to both sites and people (Tagg, 1985). The range of people and sites from which the sample is selected should be fair to the larger population. This sampling technique should allow the widest possibility for readers of the study to connect to what they are reading. In my experience maximum variation sampling provides the most effective basic strategy for selecting participants for interview studies.
Consider, for example, a study in which the interviewer wants to explore the experience of minority teachers in local teachers’ unions in urban school districts in Massachusetts (Galvan, 1990). Using the maximum variation approach, the researcher would analyze the potential population to assess the maximum range of sites and people that constitute the population.
First she would have to define what she meant by the term urban. Then she would have to determine the range of school systems in Massachusetts that fall within her definition. Within those systems she would have to decide whether she was interested in the experience of all minority teachers, those in grades K-12, or just those in some particular grade level.
In Massachusetts, local teachers’ unions are usually affiliated with either the National Education Association or the American Federation of Teachers. She would have to decide whether she was interested in studying the experience of minority teachers from both unions or from just one.
After considering the range of sites, she would then have to consider the range of people who are minority teachers and belong to local teachers’ unions. She would have to determine the relative number of male and female minority teachers, the range of ethnic groups represented, the range of subject matter they teach, their levels of teaching, and the age and experience of teachers represented in the larger population.
The above characteristics are illustrative but not exhaustive of the range of variations present in the population whose experience this researcher might want to try to understand. If the range became unmanageable, the researcher would want to limit the study, looking at, for example, the experience of one minority group in a number of locals or the experience of the full range of minority members in one or two locals. The goal would remain to sample purposely the widest variation of sites and people within the limits of the study.
In addition to selecting participants who reflect the wide range in the larger population under study, another useful approach is to select some participants who are outside that range and may in some sense be considered negative cases (Lincoln & Guba, 1985; Locke, Silverman, & Spirduso, 2004, pp. 222-223; Weiss, 1994, pp. 29-32). In the study discussed above about what it is like for a minority teacher to be a member of a local teachers’ union, it would also be useful to include some nonminority teachers who are also members of the local. If the researcher discovers through interviews that nonminority and minority teachers are having similar experiences, then the researcher will know that some issues may not be a matter of ethnicity or majority-minority status.
As another example, Schatzkamer (1986) was interested in studying the experience of older women returning to community colleges. She also decided to interview some older men who were returning to college to see in what ways their experience connected to that of the women in her sample. Selecting participants to interview who are outside the range of those at the center of the study is an effective way for interviewers to check themselves against drawing easy conclusions from their research.
2. SNARES TO AVOID IN THE SELECTION PROCESS
New interviewers may take too personally a potential participant’s reluctance to get involved. It does little good to try to persuade such a person to participate in an interview she or he would rather not do. In the face of initial reluctance, interviewers may go to great lengths to exercise persuasion only to find later the interview itself to be an ongoing struggle (Richardson et al., 1965). The interviewer must strike a balance between too easily accepting a quick expression of disinterest from a potential participant and too ardently trying to persuade a reluctant one that she or he really should participate.
Another snare is the potential participant who is too eager to be interviewed. During the contact visit an interviewer can ascertain whether the person has some ax to grind. In a contact visit Sullivan and I made to one community college, we learned that the college had just dismissed its president. The school was divided into factions: those who had worked for the president’s dismissal and those who had not. Some of the faculty we contacted were very reluctant to get involved in an interview. Others were too eager. The purpose of our study was understanding the work of community college faculty. Although it is true that academic politics are a part of that work, in this particular case the partisan politics of the campus threatened to load our study with interview participants inclined to be more like informers (Dean & Whyte, 1958; Lincoln & Guba, 1985; Richardson et al., 1965).
On occasion during a contact visit, someone would tell us we must interview a colleague who won an award and would be wonderful to talk to. Our instinct was always to avoid such “stars.” The method of in-depth interviewing elicits people’s stories in a way that shows each person to be interesting no matter how uncelebrated.
3. HOW MANY PARTICIPANTS ARE ENOUGH?
New interviewers frequently ask how many participants they must have in their study. Some researchers argue for an emerging research design in which the number of participants in a study is not established ahead of time. New participants are added as new dimensions of the issues become apparent through earlier interviews (Lincoln & Guba, 1985; Rubin & Rubin, 1995). Other researchers discuss a “snowballing” approach to selecting participants, in which one participant leads to another (Bertaux, 1981). But even if researchers use a purposeful sampling technique designed to gain maximum variation and then add to their sample through a snowballing process, they must know when they have interviewed enough participants.
There are two criteria for enough. The first is sufficiency. Are there sufficient numbers to reflect the range of participants and sites that make up the population so that others outside the sample might have a chance to connect to the experiences of those in it? In our community college study, we had to have enough participants to reflect vocational and liberal arts faculty; men, women, and minorities; and age and experience ranges. We also considered faculty with advanced degrees and without advanced degrees. In addition, we were reluctant to interview only one person in any particular category.
The other criterion is saturation of information. A number of writers (Douglas, 1976; Glaser & Strauss, 1967; Lincoln & Guba, 1985; Rubin & Rubin, 1995; Weiss, 1994) discuss a point in a study at which the interviewer begins to hear the same information reported. He or she is no longer learning anything new. Douglas (1985) is even bold enough to attempt to assess when that began to happen in his studies. If he had to pick a number, he said, it would be 25 participants.
I would be reluctant to establish such a number. “Enough” is an interactive reflection of every step of the interview process and different for each study and each researcher. The criteria of sufficiency and saturation are useful, but practical exigencies of time, money, and other resources also play a role, especially in doctoral research. On the other hand, if I were to err, I would err on the side of more rather than less. I have seen some graduate students struggle to make sense of data that are just too thin because they did not interview enough participants. Interviewing fewer participants may save time earlier in the study, but may add complications and frustration at the point of working with, analyzing, and interpreting the interview data.
The method of in-depth, phenomenological interviewing applied to a sample of participants who all experience similar structural and social conditions gives enormous power to the stories of a relatively few participants. Researchers can figure out ahead of time the range of sites and people that they would like to sample and set a goal for a certain number of participants in the study. At some point, however, the interviewer may recognize that he or she is not learning anything decidedly new and that the process of interviewing itself is becoming laborious rather than pleasurable (Bertaux, 1981). That is a time to say “enough.”
Source: Seidman Irving (2006), Interviewing As Qualitative Research: A Guide for Researchers in Education And the Social Sciences, Teachers College Press; 3rd edition.