Marc-Adélard Tremblay

Cornell University

“The Key Informant Technique :
A Nonethnographic Application.”

Un article publié dans la revue American Anthropologist, vol. 59, no 4, August 1957, pp. 688-701. American Anthropological Association.

Introduction

THE KEY INFORMANT TECHNIQUE

OBJECTIVES OF THE KEY INFORMANT TECHNIQUE IN THIS RESEARCH OPERATION

PRELIMINARY RESEARCH DESIGN

RESEARCH OPERATIONS

CONCLUDING REMARKS

REFERENCES CITED

Introduction

THERE has been extensive use of the key informant technique [1] in anthropological field work but relatively few attempts have been made to spell it out, especially from the viewpoint of its planning and its place in a structured, yet flexible, research design for data gathering. This article, which draws its material from the Stirling County Study [2], will explain why and how key informants were selected for a particular phase of the research (that of identifying the poorest and wealthiest communities of the county) with the hope that from detailed presentation of a specific case, some general principles of use can be drawn.

In this article we shall define what we mean by the technique, and then analyze its use in gathering data. This will be followed by a section on the kinds of data we hoped to discover through the use of the technique. Our research design will then be outlined and the reasons for deviating from the original design will be explored. Finally, the manner in which the operation was carried out will be described. The procedures for the analysis of the data as well as the results are not pertinent to this paper and are therefore omitted from it, but they can be found elsewhere (Tremblay 1955).

THE KEY INFORMANT TECHNIQUE

1. Definition of terms. As used here, the term "key informant" has a more delimited definition than is usual. In a traditional anthropological field research, key informants are used primarily as a source of information on a variety of topics, such as kinship and family organization, economic system, political structure, and religious beliefs and practices. In brief, they are interviewed intensively over an extensive period of time for the purpose of providing a relatively complete ethnographical description of the social and cultural patterns of their group. In that particular fashion, a few informants are interviewed [3] with the aim of securing the total patterning of a culture. The technique is preeminently suited to the gathering of the kinds of qualitative and descriptive data that are difficult or time-consuming to unearth through structured data-gathering techniques such as questionnaire surveys.

Although the emphasis is on qualitative aspects, it is also possible to get a great deal of valuable concrete quantitative data. For instance, by interviewing a saw-mill operator, one is likely to get a large amount of specific data such as the number of thousand feet of lumber sawn in a day, the number of workers required to maintain a certain rate of woodcutting, the predicted production of a piece of woodland, and so forth. This of course does not mean that qualitative data of great importance cannot be obtained in a survey. Many surveys, for instance, have open-ended questions which allow respondents to give a good deal of qualitative data, as in the Morale Survey, USSBS (Leighton 1949).

[689]

This kind of interviewing, labeled "key informant technique," is often named "the anthropological technique" or referred to as "unstructured interviewing." In our opinion, there is some objection to using either term. As for the first, despite the fact that anthropologists have made a major contribution to the development of this approach and laid down many of its principles, it cannot be considered as belonging solely to that discipline. It has been used in economics and in the political sciences, and it is also a common procedure in journalism.

On the other hand, the term "unstructured interviewing" creates the impression that the technique is of limitless plasticity and has a lack of system. As we shall demonstrate later, the technique can have structure, although it is a different kind from that used in the design and administration of questionnaire surveys. In using key informants, one chooses them strategically, considering the structure of the society and the content of the inquiry. Furthermore, in the interview itself, although the informant is given latitude to choose his own order and manner of presentation, there is a systematic attempt on the part of the researcher to cover completely the topic under analysis. When we use key informants, we are not randomly sampling from the universe of characteristics under study. Rather, we are selectively sampling specialized knowledge of the characteristics.

It should be noted that there is usually considerable difference between an anthropologically selected key informant and a statistically drawn respondent. The former is able to make comparisons between communities of the county and differentiate the poorest from the richest, mainly because this had been the criterion for selecting him. Most respondents of a questionnaire survey, being more limited in scope and knowledge, could hardly make these comparisons.

There is also emphasis on progressive restructuring of both the choice of additional informants and the content of the interviews as a result of the information gathered ; that is to say, "feed-back" [4] is an important element in the conception and operation of the method.

The term "key informant" seems to avoid the connotations of these other terms. It does not suggest any single scientific discipline, and at the same time it does imply, at least indirectly, some structuring in the selection of informants. [5] The type of interviewing may increase or lessen structure depending upon the problem.

2. A focused use of key informants. We have used key informants, in the traditional anthropological sense, during preliminary phases of the Stirling County Study and during intensive community studies. In this operation, however, our use of informants has been in the study of specific aspects of a cultural setting rather than the cultural whole usually detailed in ethnographies. The technique was in this sense very limited : the narrowness of our interest meant that we searched not for informants who might add to our total understanding of the culture, but for informants who might be expected to have specialized information on particular topics. Yet it compares with [690] ethnographic usage in that schedules are not used in the interview situation, nor are informants randomly selected as in a sample survey interview. However, it differs from the traditional anthropological technique in that a large number of key informants are selected and interviewed within a restricted framework of questions with highly focused objectives. If we were to take as our research setting a relatively unexplored culture, our interviewing procedure might be as follows : the use of ethnographic key informant technique as the first stage of investigation ; the use of the focused key informant technique at the second stage of the inquiry, to be followed, at the third stage, by sample surveys. A focused use of key informants is thus intermediate in nature. It assumes broad general knowledge of the area, but precedes the ability to choose the relevant alternatives incorporated in a well-designed sample survey.

This paper will be concerned primarily with the relatively unexplored focused use of key informants mentioned above. It is structured in the sense that the interviewer, familiar with the type of material sought from the informant, has a framework of questions in mind. This framework, which gives an idea of the type of material sought and which limits the universe to be studied, is told to the key informant at the beginning of the interview in, order to give him some orientation. If the informant's conversation is irrelevant to the topic or if he veers repeatedly from the main focus of the interview, the research worker interjects comments or questions intended to bring him back, but without forcing him to adopt a predetermined pattern of conversation. The technique is flexible in that the informant is allowed considerable leeway in regard to the content of his answers and the manner of presentation. He is encouraged to follow, by associative processes, from one thought to the other with relative freedom. A salient feature of the informant-researcher interaction is that the former is encouraged to bring out all the facts pertinent to the researcher's interest. Clues are followed and clarifications requested so that the informant's interest is continuously revived and sustained.

The technique is self-developing, since the researcher can refine his interviewing method during the course of a session, or through repeated contacts, as the amount of knowledge about the problem increases and as the ability of the informant is fully revealed. The interview process develops the informant's skills to recall facts and situations, stimulates his memory, and facilitates the expression of these recollections.

OBJECTIVES
OF THE KEY INFORMANT TECHNIQUE
IN THIS RESEARCH OPERATION

The self-developing quality of the technique and the nature of the interview data made the technique preeminently suitable for some phases of research in the Stirling County Study. A further reason for choosing it was that the size of the county, the large number of communities, and the overall research design ruled out more extensive and expensive methods of data gathering. Study planning called for comparisons of all true communities in the county on seven conditions. The key informant technique was applied to [691] one of these, poverty-affluence, with the idea that if it proved sufficiently accurate for research purposes, it could then be applied with relative ease to the remaining variables.

There were three types of data that we wanted from key informants : definitional, objective, and judgmental. These types of data were to be brought to bear on the following research objectives :

1. To develop a definition of the dimensions involved. One purpose was to evolve a conception of the nature of poverty, and its opposite, as specific phenomena in Stirling County. This is in accord with the feed-back mechanism and the process of self-development alluded to earlier.

We wanted to use the informants’ own terms for "poverty" or "wealth" instead of more abstract or more measurable terms. We feared that such terms might, by their unfamiliarity, lead the informants to unnecessarily imprecise or erroneous judgments. As a matter of procedure, the researcher gave a preliminary general assessment of what he was looking for. In response, key informants would either identify the poorest communities, in which case they would be asked to define what they meant ; or they would translate our cues into their own terms for describing poverty, after which they would be asked to identify the extremes. If we had an absolute scale, this would imply a comparison of Stirling communities against communities in general in that part of the world. But in a relative scale of the type developed here, there have to be richer communities to compare against the middle range, as well as poorer communities.

As a result of this practice, we came to define poverty as existing in communities (or subcommunities) in which the residents had no capital, few goods or possessions, low credit, no skills of economic value, and both low and irregular incomes. As the informants suggested, these conditions manifested themselves in dilapidated and unpainted houses, roofs in disrepair, untidy yards, broken windows, loose clapboards, broken steps, and shacks or big houses that had deteriorated. In the course of identifying and characterizing poorest communities, a number of key informants made comments which indicated their awareness of some of the most noted characteristics associated with poverty, e.g., lack of normal social controls, aspirations, and values comparable to those prevailing elsewhere in the county.

2. To discover boundaries of communities. Another objective of the key informant technique was to delineate the boundaries of those communities which were identified as being either the poorest or richest.

Since it was suspected that formal boundaries, such as school, postal, electoral, and church districts did not necessarily coincide with natural communities, it was important to let informants define the communities they named. The idea behind this procedure was to find groups of people, even though they might belong to two different administrative units, who regarded themselves as belonging together and as being different from those surrounding them, and who were regarded by their neighbors as being different. As it turned out, key informants mentioned a number of communities which were [692] not administrative entities and were not recorded on our maps of the county. This was especially true of the poorer sections, which were often submerged in larger and richer areas. For instance, Northwest Jonesville and The Bog were mentioned again and again as "natural areas" (an assertion which was later verified by intensive anthropological participation in these areas). They are not administrative or political units and hence do not appear on county maps as separate entities.

3. To identify extremes. Another general research objective was to identify the most extreme communities in terms of poverty and affluence, that is, the poorest and richest communities in the county. This required a relatively straightforward evaluation by the informants, although it was based on their own impressions and other subjective data, as well as on whatever objective information they might possess. The details of having informants rank communities between the extremes will be described later in this paper.

4. To increase knowledge of the problem. The final goal, which is more indirect than the others but more in line with traditional field work, was to maximize the chances of collecting relevant information not explicitly stated in the research design and, by this means, to gain further insight into some of the phenomena pertaining to the variable under investigation. [6]

PRELIMINARY RESEARCH DESIGN

1. Criteria for selection of informants. To get the data required, it was necessary to have the best possible informants. The following criteria of the "ideal" informant were delineated :

Role in community. His formal role should expose him continuously to the kind of information being sought.

Knowledge. In addition to having direct access to the information desired, the informant should have absorbed the information meaningfully.

Willingness. The informant should be willing to communicate his knowledge to the interviewer, and to co-operate with him as fully as possible.

Communicability. He should be able to communicate his knowledge in a manner that is intelligible to the social scientist.

Impartiality. As an ideal, personal bias should be at a minimum, and such biases as do exist should be known to the research worker. For instance, if the informant has a bias conditioned by his class position, this should be known to the interviewer so that its effects can be properly appraised.

Of these five criteria of eligibility, only role in community can be determined in advance. The other qualifications are apt to be largely matters of personality, rather than positions in the social structure. Once individuals performing key roles in the economic structure are detected, the other four criteria serve as a screening device for separating the "good" from the "poorer" informants. This means that, after having prepared an ideal list of informants [693] on the basis of their roles in the community, we could expect to make some changes as a result of personal contact and appraisal. It was also anticipated that repeated contacts with informants might lead to the best ones being singled out for more attention.

For judging the information provided by the informants, and in fact the informants themselves, the following criteria were considered important : internal consistency, productivity, and reliability. These criteria are preliminary to checks against outside standards ; e.g. census materials, surveys, intensive field work in the locations. Let us spell out these preliminary checks.

Internal consistency. This is a necessary, although not sufficient, condition for accurate information. Such consistency can be checked in the course of data-collection and analysis of each informant. Furthermore, there should be a cogent explanation for any specialized knowledge which key informants exhibit in the interview situation. This is especially true when the richness of detail goes beyond what one would expect. For example, one individual was particularly detailed in his accounts of the economic and social relations of families in one section of the county. Upon being asked why he knew all these facts, he cited his participation in numerous fund-raising campaigns, his career as a newspaper reporter, and his work in local government committees concerned with taxes, education, and police administration.

Productivity. Productivity implies the ability to tell a lot about the problem. In our case it meant that the informant knew a large number of communities and a great deal about their economic structure.

Reliability. In a technique of this kind, cross-comparison is feasible and should be utilized as much as possible during data collection. This will give some indication of reliability and reveal areas of discrepancy where more intensive interviewing may be needed.

It is worth noting that in every sizable community there are one or two individuals with particular skills as informants. We have come to designate them "natural observers." This term was suggested by Alexander H. Leighton, from whose experiences as a field worker many of our observations are derived. These people have been lifelong students of the human scene. They are interested in the behavior of their fellow men, observe the development of institutions, and often speculate and make inferences about both. Usually, there is no one in the community where they live with whom they can exchange these interests, and the appearance of the social scientist seems to afford them considerable satisfaction. The qualities of the "natural observer" appear in a variety of roles. Such people sometimes have very limited horizons, as in isolated farms and small villages. More often, however, they occupy positions from which they can derive a broad knowledge of human affairs, e.g. police magistrate, municipal clerk, teacher, or doctor in rural districts.

2. Preliminary selection of informants. On the basis of formal role participation, a preliminary list of informants was developed. Our choice of roles was [694] determined both by the nature of the information sought and by the political structure of the county. The first, the distribution of poverty and wealth, meant choosing informants whose roles might provide them with wide and accurate knowledge of financial conditions in the county. Among these were the more obvious roles of bankers, large-scale employers, and local government and welfare personnel, as well as such roles as those of newspaper reporters and doctors, whose work might be expected to lead to wide general knowledge of the county. The second determinant was the fact that the county is divided into two dissimilar municipalities. To keep our data symmetrical, we imposed the requirement of aiming for an equal number of informants, as well as comparable role-representation, in each sector. The number of people in the county who filled these two qualifications of role-eligibility and symmetry are indicated in the following table.

RESEARCH OPERATIONS

1. Deviation from the preliminary design. Deviations were introduced into the design as it was applied in the field. There were a number of causes for deviation : the overlapping of roles ; the application of the last four criteria for informants to those selected on only the first criterion (role-participation) ; the discovery of individuals who fulfilled these four criteria but who did not occupy formal positions that suggested their special knowledge ; and limitations inherent in the field situation.

The overlapping of roles. Some individuals occupied more than one of the roles selected as a point of departure. Some of the saw-mill operators and cooperative managers, for instance, were also municipal councilors. Thus, where the symmetrical design called for two separate interviews, the field operation yielded only one.

Lack of knowledge. This factor was the basis for eliminating many individuals from the original list. A short contact with people occupying some of the formal [695] roles made it evident that a prolonged interview would yield little valuable information.

Discovery of new informants. In the course of contacting and interviewing people occupying the listed roles, some individuals suggested others whom they considered particularly well qualified as informants. In a number of cases, contact was made and relevant data were collected.

Intensive versus extensive interviewing. A few informants were highly productive and exceptionally well qualified by all the criteria mentioned earlier. In order to get the full detail of their knowledge, it was essential to interview them more often than was anticipated. Particularly in the urban center of Bristol, numerous interviews of three and four hours each had to be secured, and one key informant was interviewed at regular intervals during three months. Since time was limited, this made it impossible for us to contact some potential informants listed, although we aimed at interviewing a maximum number.

* Altogether there were twenty-eight key informants who occupied a total of forty-one major economic roles, The difference between the two numbers results from duplicate roles.

Personality factors. As with lack of knowledge, a few individuals had to be left out mainly because of personality factors that interfered with, or made impossible, communication between field worker and informant.

[696]

Operational inconveniences. This refers to practical factors in the field. One example was the impossibility of interviewing two of the bankers in the largest town in the study area because it was thought necessary to get permission from the companies’ head offices in Montreal. By the time this could have been accomplished, the research as a whole would have passed beyond the deadline for this particular operation.

In brief, the selection of informants was not based on representativeness of age, sex, and locality of residence. The latter would have been important if these individuals had been randomly selected respondents rather than judgmentally selected key informants. The selection was based almost exclusively on intensive knowledge of many communities in the county and ability to impart that knowledge to the interviewer. The symmetrical design was of great utility in maximizing the chances of locating individuals who combined a high degree of knowledge with the ability to communicate it accurately. In short, it was a device for finding "good" informants by first looking into the formal roles which they were likely to occupy.

2. Management of the interview. As noted earlier, many of the informants had been previously interviewed by members of the Project, and therefore had a fairly sophisticated knowledge of the general nature of our work. The few who did not know the Study's goals and activities were given a general introductory explanation and a printed brochure as means of orientation.

However, all key informants needed explanations on the nature of the immediate task. A standard, yet flexible, procedure was developed by the two main interviewers in order to maximize consistency and to get comparable qualitative data.

As the first step, the Study's interest in grading and comparing communities of the county from the standpoint of material wealth was discussed. The key informant was then given a map and asked to consider the communities he knew best, and to rate them on a continuum of material wealth.

It was decided to use a map rather than a check-list of the place names as a memory aid, because of the quality of the data we hoped to procure. A checklist would probably have resulted in maximally standardized procedures (i.e., entire coverage, same order of presentation of communities, etc.). However, it would probably have minimized the informant's sense of freedom to concentrate his discussion on the communities he knew best, regardless of instructions. He might have felt obliged to discuss all names, whether he was well informed on them or not.

The use of the map would avoid this disadvantage. Used only as a visual aid to stimulate the informant to organize his material himself, it would encourage his thinking on those very communities he did know best. The disadvantage to this system was that the informant was likely to overlook some communities. This could be overcome to some extent by having the interviewer ask directly about any areas of the county which the informant had not discussed, after the informant had given detailed information on the communities he knew best.

[697]

After the informant had a chance to examine the map, he was asked to pinpoint (1) the poorest and (2) the richest communities. The order was to sharpen the informants' sense of contrasts between the extremes of poverty and affluence. After the informant had enumerated all the communities which he tagged either poorest or richest (and the reasons therefore), he was then asked to single out (1) all the poorer than average communities which were not so poor as the poorest, and (2) all the better than average communities which were not so affluent as the richest. After additional queries had been made, communities which had not been rated were therefore either unknown to the informant or considered as average. In such cases, communities which the informant considered average or did not know, were often undifferentiated by this procedure.

After all the ratings had been completed for these four categories, the informant was asked to rank-order the communities he had mentioned within each class of wealth. For instance, if a respondent had mentioned Loomervale, The Bog, and Monkeytown as belonging to the poorest class, he was then asked to rank-order these three from poorest to least poor. In numerous instances, however, informants felt that they could not make such refined distinctions ; they were not pressed further. Because rank-orderings within wealth categories were incomplete, it was impossible to develop a method which could refine further the within-class-rating of informants.

In accord with our aim of defining poverty and wealth in the local idiom, we encouraged informants to explain their reasons for rating the communities as they did. We also encouraged them to delimit and describe the places they rated, since we were interested in isolating all true communities in the area.

We tried to record the interviews as fully as possible. Colored pencils were supplied and informants were encouraged to use them to spot the communities they discussed. Such maps were kept as part of the interview record. Informants were interviewed to the fullest extent compatible with their knowledge. Some required only one interview to exhaust the relevant information, while others needed more interviews. Extensive notes were taken during the interviews, and on some occasions total recordings were made.

CONCLUDING REMARKS

Although this paper deals with the illustrative case of poverty, this procedure was repeated for other sociocultural factors similarly relevant to our main problem. On the basis of these focused key informant operations, we were able to gather the information necessary for the design of a sample survey to be used in the study area as a whole, and for the preliminary selection of focus communities for intensive analysis with both "structured" and "non-structured" techniques of interviewing and observation. Thus, the technique not only provided us with the information essential for the refinement of the overall research design, but it also dictated the steps whereby its validity could be checked through comparison with the findings of subsequent research operations.

[698]

At a later date we hope to publish the results of this validity check for the technique described in this paper, and to set forth some comparisons of results achieved with this and with other research tools in the course of our study.

[699]

REFERENCES CITED

References marked with an asterisk are also found in the Selected Bibliography.

*KLUCKHOHN, CLYDE

LEIGHTON, ALEXANDER H.

*MADGE, JOHN

MERTON, ROBERT K.

*NADEL, S. F.

OPLER, MORRIS and RUDRA DATT SINGH

OSGOOD, CORNELIUS

*PAUL, BENJAMIN D.

*ROYAL ANTHROPOLOGICAL INSTITUTE

SPICER, EDWARD H. (ed.)

TREMBLAY, M.-ADÉLARD

WIENER, NORBERT

YANG, MARTIN

SELECTED BIBLIOGRAPHY

BARTLETT, F. C.

BECKER, H. S.

BENNETT, J. W.

BOAS, FRANZ

[700]

CAWBELL, DONALD

CANNELL, CHARLES F. and ROBERT L. KAHN

CHAPPLE, E. D.

DEAN, JOHN P.

DOTSON, FLOYD

DUBOIS, CORA

FESTINGER, L. and D. KATZ (eds.)

GARRETT, A.

HALLOWELL, A. IRVING

HERSKOVITS, MELVILLE

JAHODA, M., M. DEUTSCH, and STUART W. COOK (eds.)

KLUCKHOHN, CLYDE

MACCOBY, ELEANOR AND NATHAN MACCOBY

MADGE, JOHN

MALINOWSKI, BRONISLAW

MEAD, MARGARET

MERTON, ROBERT K.

MERTON, R. and P. KENDALL

NADEL, S. F.

OSGOOD, CORNELIUS

PAUL BENJAMIN D.

RADIN, PAUL

REDFIELD, ROBERT

ROGERS, CARL R.

ROYAL ANTHROPOLOGICAL INSTITUTE

SEWELL, WILLIAM H.

STAVRIANIS, B. K.

STEWARD, JULIAN H.

WARNER, L. and P. S. LUNT

WHITING, JOHN and OTHERS

WHYTE, WILLIAM : F.

WORMER, M. and C. SELLITZ