Interviewing in Social Sciences Research Paper

1. Situating The Interview

Interviews may be conducted in a variety of ways and in a variety of situations. Inevitably they are socially and methodologically embedded in larger contexts. Relatively little research has been done either on the impact of different social situations and settings on the conduct of interviews, or the impact of varying situations on the reliability and validity of interview results. The social setting may range from the privacy of the home to the office to the public ‘person on the street,’ and the social context may range from a highly private, anonymous, or confidential one-on-one interview to one involving the presence of strangers, or intimate family and friends as co-respondents. Re-search has shown that in contrast to the one-on-one interviewer respondent situation, the presence of others tends to produce more socially normative responses (Lever 1981). Much more research is needed on the impact of these varying social contexts.

We must also situate interviews in the different methodologies or styles of research used by social scientists. For heuristic purposes we can identify four major styles of research—experimental, survey, field-work, and nonreactive research. The interview, though found in all four methods, is particularly important in survey and field research and these may be thought of generally as representing quantitatively oriented vs. qualitatively oriented interview styles, respectively. The quantitative qualitative distinction itself encompasses a variety of comparisons that may range from more philosophical and epistemological issues such as objective vs. subjective knowledge, to more grounded and specific issues of questionnaire construction and closed vs. open answer formats. A central distinction running throughout these comparisons is the question of the nature of control over the interview format and content.

2. A Note On Control Over The Interview—Form And Content

Different interview formats may be arrayed on a number of different continua but of particular significance is the question of varying degrees of control. At one extreme the interviewer exerts full control over the situation and at the other the interviewer seemingly minimizes control, relinquishing more control over the situation to the respondent. There is a zero-sum quality of control in the relationship between inter-viewer and respondent. The case of maximal inter-viewer control is exemplified by the ‘forced choice’ survey interview where not only the format of the question is standardized (stimulus) but the possible answers of the respondent are standardized as well (response). The interviewer ensures a constancy of routinized control in order to generate highly reliable data. Concerns over such control are often associated with surveys where the interview is used as a tool to generate standardized data from a sample of numerous units for the purposes of aggregate analysis that may be generalized to some larger population.

The other extreme, where the interviewer exercises minimal control, is seen for example in the Rogerian psychoanalytic interview. Here the respondent is ‘in control’ as to the content of the interview and the interviewer is trained to be ‘nonreactive.’ It should be noted, however, that even here the interviewer is exercising control over the format or form of the interview, and that both the forced choice survey interview and the nonreactive interview (Rogers 1945) are interaction situations far removed from the ‘natural’ form and content of everyday personal interaction. The field interview of participant observation is, by contrast, a situation often closer to the give and take of everyday ‘normal conversation’ (Seidman 1998). The respondent exercises a degree of control over both the content and the form, but the interviewer, through judicious use of probes, can often guide the interview into topic areas of interest to the researcher. This is the case of the ‘focused’ or ‘topical’ interview (Merton 1956). From these considerations one may construct a typology of control which the researcher exercises in the interview situation:

Typology of interviewer control in different styles

of research interviews

                        Control over content

                        Hi                                                        Low

Hi                                 survey questionnaire                             nondirective

                                                                                    psychoanalytic

                        Control over form

Low semifocused in-depth conversational fieldwork

These different interview formats are often associated with survey vs. fieldwork but extend to a variety of methods. A multimethod strategy which emphasizes different strengths and weaknesses of different methods or styles of research nonetheless values the interview as a central data collection tool that can simultaneously provide quantitative and qualitative data, and even be used in multimethod experimental strategies. For example Schuman et al. (1985) incorporated an experimental design into a survey by having interviewers systematically vary the format of questions asked of different respondents.

3. The Theory Of The Spring

Some have suggested that all measurement involves interaction (Babbie 1998)—interaction with the external world in some form or another in order that information about that external world may be transferred to an observer. Theories of measurement take into account a detailed exposition of the measurement process itself as constitutive of the data generated. In the prosaic act of weighing a bunch of bananas in the produce section of a grocery store, for example, one is operating on an assumption about how the spring in the scale responds to varying masses of bananas subjected to the law of gravity in moving the dial in a consistent and credible way. One operates on the assumption that some scientist or engineer has worked out in detail the physical elasticity of the steel spring at the heart of the scale’s measurement process. In the interview this interaction is particularly noteworthy since it is social interaction—usually between two people—and the corresponding theories of measurement most applicable are theories of social interaction itself. In this section we will briefly apply four distinct theories of social interaction to the interview to see the varying perspectives they provide on this all-important data-gathering tool. The four theories of social interaction are: (a) role theory, (b) exchange theory, (c) conflict theory, and (d) symbolic interaction.

4. Role Theory

Role theory begins with a set of normative expectations that are presumed to define particular positions or statuses in social structure and their corresponding roles or behaviors in interaction with others. The roles or statuses most clearly central here are those of ‘interviewer’ and ‘respondent’ themselves. How well one knows these roles, that is, knows the normative expectations for behavior associated with the respective roles, is a function of one’s prior experience and knowledge gained either first hand/or through vi-carious observations—in the media, through cartoons, conversations, or the classroom. To be ‘trained’ as an interviewer is to learn a set of normative expectations about how one should interact with a respondent. The simplest norms are of ten those that are ‘taken for granted’ such as who asks the questions and who gives the answers. Sometimes there are widely accepted norms defining the ideal interviewer role such as—one should not reveal personal information that might thereby bias the respondent’s responses (Gordon 1975). However, even these commonplace widely shared norms may come into question as there may be conflict among the different goals for the interview. For example, one goal may be to get as much revelatory information as possible vs. another goal which is to not bias the respondent’s answers. The different goals of the interview (unbiased data vs. more fully revelatory data) may come into conflict and suggest different contradictory norms, for example, about how much personal information the interviewer should reveal to the respondent in the interview situation.

Role theory asserts that the norms governing interaction are there for the purpose of realizing specified goals out of the interaction. In short, normative behavior is goal oriented. The norms governing the interviewer are likewise oriented to maximizing certain qualities (or values) of the data—for example, norms that call for a common stimulus to achieve validity across respondents, or other norms that emphasize supportive interaction to produce full and complete responses.

A common application of role theory to the interview situation is a concern with bias introduced by other social statuses in a person’s status set—the most obvious and frequently researched being the ascribed and highly visible social statuses of race, gender, and age. These statuses are often considered to be especially important when questions in the interview touch on topics closely related to them. Much research has been conducted on the effects of these statuses on interview outcomes precisely because they are considered to be master ascribed statuses that are both readily observed and ubiquitous in all interactions. They are often attributed even in telephone interviewing and not just in face-to-face situations.

One of the earliest findings of interviewer effects by race, dating to World War II, was that answers to questions about race relations were strongly impacted by the races of the interviewer and respondent (Hyman et al. 1954). African-Americans reported lower levels of satisfaction with race relations to African-American interviewers than to Caucasian interviewers. Caucasian respondents reported lower levels of acceptance of African-Americans to Caucasian interviewers than to African-American interviewers. These results have been demonstrated repeatedly since then in studies by Schuman et al. (1985) and Anderson et al. (1988). The latter study found that blacks interviewed by whites were much more likely to express warmth and close-ness toward whites than blacks interviewed by blacks. It is important to note that more limited effects of race of interviewer and respondent have been observed on studies of other topics, unrelated to race.

Another interviewer status examined in some detail has been gender. Earlier research has generally shown that the gender of the interviewer has no effect on responses to survey questions except when the content of the questions relate to sexual behavior or gender related issues (Clark 1952, Hyman et al. 1954). Research by Kane and Macaulay (1993) found that both male and female respondents express more egalitarian gender related attitudes or greater criticism of existing gender inequalities to female interviewers. Furthermore, male respondents offer significantly different responses to male and females interviewers on questions dealing with gender inequality.

Age, as the third of the most visible ascribed status characteristics, has also been found to have only limited interviewer effects. School age subjects have shown different responses to older vs. younger inter-viewers for questions about peer influences (Ehrlich and Riesman 1961). Younger interviewers obtained slightly more peer-oriented and less adult-oriented answers than older interviewers. With respect to age of respondents, research has shown that telephone surveys tend to under-represent older respondents, but response distributions do not vary by telephone vs. face-to-face interviews across age categories (Herzog et al. 1983). As with race and gender, age appears to have little effect on interview responses except when the topic being covered is directly related to the status itself.

5. Exchange Theory

The central assumption of exchange theory is that all interaction is based on ego’s attempt to realize a ‘social profit’ by minimizing the cost of one’s own actions and maximizing the benefits from alter’s actions. The operating principle is that for interaction to continue, the norm of reciprocity must be followed such that the ratio of two parties’ costs to benefits should be equal. If we view the interview as an exchange, it is relatively easy to define the benefits to the researcher; they are the data provided by the respondent which are his or her costs in time and effort in answering the questions. It is often more nebulous to define what the benefits to the respondent are, that is, the costs incurred by the interviewer that directly benefit the respondent. Of course when respondents are being directly paid by the interviewer for their time, the benefits are clear and unambiguous. Market researchers, for example, sometimes establish panels of randomly selected respondents who are paid for their time in a variety of repeated surveys. In short, they are ‘professional respondents’ for whom the exchange is a pure labor market transaction. Exchanges in kind may also be provided such as a lunch or dinner, a free gift, or a chance at a drawing. More often than not, however, the benefits are assumed to be some nebulous internal psychic reward that can include everything from the opportunity to vent one’s opinions on a particular topic, that is, make use of a willing listener, to an increased sense of self-worth engendered by volunteering to help someone out, especially when the interviewer may have already appealed to one’s ‘specialness’ by indicating that they have been ‘selected’ to participate. Altruistic appeals also redound to one’s sense of self-worth as when one is solicited as a respondent to provide data that may be useful in developing new policies and programs or evaluating existing ones even if one is not seen to be a direct beneficiary of them. Even the most abstract altruistic appeal—that by responding one is ‘contributing to knowledge’—may at times work as a sufficient psychic reward.

What the exchange perspective highlights is that ‘time spent’ by the respondent is a very real but variable cost that may influence their willingness to be interviewed in the first place, and their loquaciousness even if an interview is granted. The preciousness of ‘time spent’ doing an interview is differentially valued from one potential respondent to another (for ex- ample, the working mom vs. the elderly retired nursing-home resident). Such calculations are highly likely to influence differential response rates in a selective or biased way. Not only are these calculations likely to vary from one set of statuses to another, but they will likely vary even for the same individual under different circumstances and in different situations, such as time of day or day of week or season of the year, or at home, shopping, or workPol 1992).

The cost of an interview is not simply one-sided, for the interviewer’s time is also a very tangible cost, and one that must be taken into consideration often in a direct monetary way as when one is hiring interviewers or relying on a survey supplier. Such costs may have direct bearing on the nature and quality of the data gathered, for they can affect the length of the interview forcing real choices in research design as to sample frame and size, or number of topics explored and questions asked.

In short, from an exchange perspective the interview is a cost/benefit transaction. The calculus of these varying costs to the interviewer researcher must be weighed against the value of the data to be generated. In the design of surveys, for example, a distinction is often made between open-ended vs. closed-ended questions, and the former generally prove to be more difficult and more expensive to code and analyze than those from closed-ended questions. However, Geer (1991) found in an experiment that the open-ended comments are very effective in measuring the important concerns or ‘salient’ issues of respondents. In short, it appears to be cost considerations as much as validity issues that have led to the dominance of forced choice questions in survey research.

Costs and benefits of the interview also form a part of the calculus of the ethical considerations involving human subjects. Costs to the respondent slide over into a variety of ‘risks’ that may result from the disclosure of certain data and the loss of privacy. Anonymity or confidentiality may be required to forestall those costs. The idea behind ‘informed’ consent is that this calculus should be made by the subject respondent based on a full disclosure of risks and benefits. Research in which there are substantial risks can only be justified by substantial benefits that outweigh these costs, and above all it is up to the respondent to make this determination.

We should be aware that the utilities (value) of a respondent’s cost benefits may differ from those of the researcher and one is confronted with incomensur-ability which may work to the advantage of the researcher. For example, an elderly retired single respondent may have a surplus of free time which is of low cost to give in answering a lengthy questionnaire, while to the highly trained interviewer this time is a valuable costly commodity in the research budget. Such incommensurability may at times work to the detriment of the interviewer when the inability to reach a common understanding of the relative value of information or time leads to an unwillingness to participate in what is considered to be a ‘worthless’ activity. (We note that this is one of the points of the symbolic interactionist perspective outlined below which involves negotiating not only a common value but a common understanding of the meaning as well as the value of various social goods.)

6. Power Theory

The power or conflict perspective directly addresses the insight that much social interaction—and also the interview—involves differences in power among the parties. To be sure, power or conflict theory may be closely linked to either role theory (as when, for example, one talks about the differences in power that are present between the status of a male and a female in interaction) and also to exchange theory (as when the costs to participating in the interview are different between the interviewer and the respondent, or the benefits that one can provide to the other are different and one or the other party uses those differences in resources to extract what they want from the other). What the power or conflict perspective stresses is the reality that interviewer and respondent often have different degrees of power relative to one another and these may influence the nature and quality of the data generated in the interview.

Let us first address situations in which the inter-viewer has more power. We will use power in its broadest meaning to encompass the most subtle uses of prestige to the blatant uses of physical coercion. The latter is rare if ever present but the former is quite ubiquitous. The imprimatur of an individual’s title ‘Doctor’ or ‘Professor,’ or an institution’s stature ‘X University,’ or ‘Y Agency’ are often used to impress potential respondents as to the importance of the research and the desirability of their participating. In the actual conduct of the interview itself power may be reflected in who makes the effort to initiate the contact or who travels to whom to conduct the interview. Being on one’s home turf would seem to imply more power and control over the situation. A notable exception is the case in which elites may avoid being interviewed in certain situations (coming to their homes) in order to maintain a heightened degree of privacy. Researchers have sometimes strategically relied on the interviewer shifting out of the role of interviewer to becoming an observer and recording the presence or absence of such things as a piano, types of pictures and paintings hung on walls, and the use of curtains or drapes (Laumann and House 1970). This is an example of combining methods in a multimethod strategy (Brewer and Hunter 1990) to provide validity checks surrounding lifestyle and status and class variables.

At times respondents may sense a threat from the interview situation that reflects a potential power differential between them and their interviewers. Potential threats may apply to divulging information concerning content areas that include socially undesirable and even illegal behaviors and could lead to not reporting or under-reporting various forms of deviance and even criminal behavior ranging from drug use to official ‘cheating’ (Bradburn et al. 1979). Implicit threat could also lead to over-reporting socially desirable behavior such as voting or charitable giving. There is some limited evidence that interviewer characteristics are related to threat. Respondents are willing to report socially undesirable behavior such as drug use to interviewers who are perceived to be more like themselves (Johnson et al. 2000).

The most widely used procedure to reduce threat is to eliminate the interviewer entirely or as far as possible from the data gathering process. In some cases this means group administration of the interview in a public or quasipubic setting but where responses are private, such as a school classroom setting. Even if the interviewer is present, it is possible that the threatening questions may be self-administered. Maxi-mum privacy is obtained when the threatening questions are on a computer and the respondent reads questions from a screen. The respondent enters the answers directly into the computer and the interviewer is completely unaware of what answers are given (Tourangeau and Smith 1996). It must be recognized that the use of self-administered methods reduces but does not eliminate respondent misreporting due to the power differentials in the interview situation and the potential threat of disclosure. It is precisely why the use and assurance of anonymity or confidentiality are not only ethical considerations but themselves may have a direct and positive impact on the validity of data generated through the interview.

Power is implicit in the very asymmetry of who asks questions and who answers them. The benefit is seen to be asymmetrical as the powerful ask questions of, glean information from, and acquire data about the less powerful. Much more rarely does information flow in the reverse direction—from the powerful to the powerless. Knowledge is power and its differential distribution is a means of obtaining and maintaining that power (Hunter 1993).

Interviewing of elites is a situation in which the respondent often has more power than the interviewer. Such control may be reflected first in limiting the ability to gain access to begin with and continue through the interview situation with control over the content, format, recording, and even the use of interview data (Thomas 1993). Although interviewers in their comportment should, of course, show respect to all respondents, an added degree of deference might be shown to those having more power or higher status than the interviewer. This deference could easily affect the likelihood of the interviewer’s continuing to probe in finer and finer detail into the lives of the respondents, thereby limiting the quality and depth of the data.

The power differential enters as well into ethical considerations of the interview, and this is why interviews must be voluntary, since the counter case involving any form of coercion, no matter how mild, may be deemed unethical. For example, if a welfare recipient is being interviewed for research purposes, they might agree to participate in the interview if it is closely linked to the intake interview used to determine welfare elgibility; or, a person in a public setting may agree to be interviewed out of a sense of peer pressure (e.g., students in a classroom setting). It is imperative that researchers be fully cognizant of and sensitive to the power relations involved in the interview as these may not only affect the quality of the data but reflect larger social processes of which the interview itself is a microcosm.

7. Symbolic Interaction

The interview situation is not one which is necessarily clearly defined and unambiguous, especially on the part of respondents. Therefore, one of the key activities of the good interviewer is to negotiate with the respondent in defining the appropriate behaviors and what can be expected in the interview situation. Drawing on the symbolic interactionist framework (Goffman 1959), the interview may be considered to be a situation in need of mutual definition through the interaction of the interviewer and respondent. At a minimal level is the idea that the interviewer is supposed to ask questions and the respondent is to give answers. But the fine nuances of this somewhat ‘artificial’ interaction are a negotiated reality in which the interviewer has the task of ‘educating’ the respondent into the appropriate behavior. For example, a too loquacious respondent might be interrupted in the course of an answer and asked to respond more simply—‘strongly agree, agree, neutral, disagree, or strongly disagree.’

The negotiated definition of the situation draws on the past experiences of both interviewer and respondent. The idea of interviewer training focuses on behaviors used to aid in an appropriate definition of the situation such as what to say to gain access, how to present self for gaining agreement to participate, how to develop and maintain rapport and how to terminate an interview, and of course, how to avoid biasing respondents (Billiet and Loosveldt 1988). Some respondents may be familiar with the interview situation from having participated in them previously, while others may have been exposed to the situation vicariously through the media. Some may in fact be ‘professional respondents,’ that is, those paid to participate in ongoing panel studies in which they have been more thoroughly trained or educated into the respondent’s role.

Expectations of the interviewer about the interview process have also been found to influence the nature of the respondents’ participation. Some interviewer expectations are formed before the interviewer conducts any interviews, for example, expectations about how difficult or threatening certain questions may be to respondents; while other expectations may develop in the course of an interview as the interviewer begins to form a picture of the respondent through their prior responses, and yet other expectations may emerge as the interviewer completes multiple interviews and begins to form a patterned set of expectations for emerging ‘types’ of respondents. For example, inter-viewers who expected that questions about threatening behavior would be easy to ask received higher rates of responding than interviewers who thought the questions would be difficult to ask (Bradburn et al. 1979). It would appear that interviewer behavior in some way reflects expectations and that behavior calls forth different responses on the part of subjects.

Ambiguity in social interaction is a key problematic in the symbolic interactionist framework. Much interaction is seen to be specifically oriented to the reduction of ambiguity in an attempt to arrive at a mutually agreed upon definition of the situation. The issue of ambiguity is also significant in the interaction between interviewer and respondent. If a respondent should ask quizzically, what is meant by a question, often interviewers are simply instructed to repeat the question without further elaboration or clarification of the question so as no to differentially bias one respondent vs. another. On the other hand, if a respondent gives an answer that is limited, partial, or ambiguous, the interviewer is often instructed to ‘probe’ to get greater clarification. In a classic lab-oratory experiment, Hyman et al. (1954) demonstrated that interviewers faced with an ambiguous response coded the answer on the basis of previous answers in the questionnaire. Respondents were asked a series of questions about whether the United States should actively intervene or should isolate itself from world affairs. When the response was ambiguous, those who had previously given unambiguously interventionist answers were coded as giving an interventionist answer, while those who had previously given isolationist answers were coded as giving an isolationist answer. In this experiment interviewers were not allowed to probe the ambiguous answer for clarification.

The first phase of an interview involves negotiation in securing the willingness of the respondent to participate. By negotiating a formal ‘commitment’ from respondents in the form of a signed ‘pledge,’ to answer ‘fully’ and ‘truthfully’ has been found to increase the degree of cooperation by respondents and the completeness of their responses (Cannell et al. 1981). A further mechanism for increasing cooperation is to flexibly tailor the appeal to participate to each individual respondent rather than using a standardized approach. Such individually tailored appeals generate higher levels of cooperation and also experienced interviewers are more likely to obtain such higher rates of cooperation (Groves and Couper 1998).

8. A Multimethod Strategy

Interviews in their various forms and formats will continue to be a primary tool in the social scientist’s toolkit for generating reliable and valid data. Direct comparisons of the interview to other tools for data generating, such as content analysis of media, or conversational analysis, the use of systematic observation of behavior, or the analysis of archives can only further the understanding of the strengths and limits of the interview in a multimethod strategy. Some but far too little research has been concerned with such comparisons. One example is Janet Lever’s (1981) multimethod study of children’s play. She found that interviews with kids produced more gender stereo-typed responses to the nature of boy’s vs. girl’s play than observations of their behavior by the researcher or diaries kept by the kids themselves. Throughout the above discussion the multimethod strategy has been mentioned at various points and briefly noted in passing. Bringing the multimethod strategy to the foreground whether narrowly focused on measurement or more broadly on the research process as a whole (Brewer and Hunter 1990) can only lead to a more refined use of the interview in social science research. The shape and function of the interview as a tool can more clearly be delineated as it is compared to these other tools in the same research project. The centrality and the ubiquity of the interview requires much more systematic research fully exploring the social, situational, and methodological dynamics of the interview process. Whether viewed from a post-modern perspective of a greater need for reflexivity within the social sciences or a positivist perspective of understanding the sources of bias and the need to reduce measurement error, much more research on the interview itself is needed. A few lines of inquiry and understanding have been developed as noted, but a great abyss of ignorance remains to be plumbed.

At various points throughout this discussion I refer to data being either ‘gathered’ or ‘generated.’ I fully recognize the implications of these two distinct concepts as the former privileges an older classical positivism while the latter speaks to a more agency oriented social constructionist view of science. The distinction will itself become part of our nuance discussion of the interview.

Bibliography:

1. Babbie E 1998 The Practice of Social Research, 8th edn. Wadsworth, Belmont, CA
2. Billiet J, Loosveldt G 1988 Improvement of the quality of responses to factual survey questions by interviewer training. Public Opinion Quarterly 52: 190–211
3. Bradburn N, Sudman S et al. 1979 Improving Interview Method and Questionnaire Design. Jossey-Bass Publishers, San Fran-cisco
4. Brewer J, Hunter A 1990 Multimethod Research: A Synthesis of Styles. Sage Publications, Newbury Park, CA
5. Cannell C, Miller P, Oksenberg L 1981 Research on interviewing techniques. Sociological Methodology 12: 389–437
6. Clark R A 1952 The Projective Measurement of Experimentally Induced Levels of Sexual Motivation. Journal of Experimental Psychology 44: 391–9
7. Geer J 1991 Do open-ended questions measure ‘salient issues’? Public Opinion Quarterly 55: 360–70
8. Goffman E 1959 The Presentation of Self in Everyday Life. Doubleday Anchor, New York
9. Gordon R 1975 Interviewing: Strategy, Techniques, and Tactics. Dorsey Press, Homewood, IL
10. Groves R, Couper M 1998 Nonresponse in Household Interview Surveys. John Wiley and Sons, Inc., New York
11. Herzog A R, Rodgers W, Kulka R 1983 Interviewing older adults. Public Opinion Quarterly 47: 405–18
12. Hunter A 1993 Local knowledge and local power: notes on the ethnography of lcoal community elites. Journal of Contemporary Ethnography 93: 36–58
13. Hyman H, Cobb W, Feldman J, Hart C, Stember C 1954 Interviewing in Social Research. University of Chicago Press, Chicago
14. Johnson T, Fendrich M, Shaligram C, Garcy A, Gillespie S 2000 An evaluation of the effects of interviewer characteristics in an RDD telephone survey of drug use. Journal of Drug Issues 30: 77–101
15. Kane E, Macaulay L 1993 Interviewer gender and gender attitudes. Public Opinion Quarterly 57: 1–28
16. Laumann E, House J 1970 Living room styles and social attributes. Sociology and Social Research 54: 321–42
17. Lever J 1981 Multiple methods of data collection: a note on divergence. Urban Life 10: 199–213
18. Merton R 1956 The Focused Interview: A Manual of Problems and Procedures. The Free Press, New York
19. Pol L 1992 A method to increase response when external interference and time constraints reduce interview quality. Public Opinion Quarterly 356–60
20. Rogers C 1945 The non-directive method as a technique for social research. American Journal of Sociology 51: 143–52
21. Schober M, Conrad F 1997 Conversational interviewing. Public Opinion Quarterly 61: 576–602
22. Schuman H, Steeh C, Bobo L 1985 Racial Attitudes in America: Trends and Interpretations. Harvard University Press, Cam-bridge, MA
23. Seidman I 1998 Interviewing as Qualitative Research, 2nd edn. Columbia University Teachers College Press, New York
24. Suchman L, Jordan B 1990 Interactional troubles in face-to-face survey interviews. Journal of the American Statistical Association 85: 2322–53
25. Suchman L, Jordan B 1991 Validity and collaborative construction of meaning in face-to-face surveys. In: Tanur J (ed.) Questions about Questions: Inquiries into the Cognitive Bases of Surveys. Russell Sage Foundation, New York, pp. 241–67
26. Sudman S, Bradburn N 1974 Response Effects in Surveys: A Review and Synthesis. Aldine Publishing Company, Chicago
27. Thomas R 1993 Interviewing important people in big companies. Journal of Contemporary Ethnography 22: 80–96
28. Tourangeau R, Smith T 1996 Asking sensitive questions. Public Opinion Quarterly 60: 275–304