Reason Analysis

views updated

Reason Analysis

Assessment of cause

When to apply reason analysis

Models of action and accounting schemes

Designing a reason analysis

Strategies for assessing cause

The validity of reason analysis

BIBLIOGRAPHY

Reason analysis is a set of procedures used in survey research to construct causal explanations for the actions, decisions, or intentions of individuals (Lazarsfeld 1935). It involves framing and asking questions, as well as coding and analyzing replies of respondents. Typical actions that have been subjected to reason analysis include buying a product (Kornhauser & Lazarsfeld 1935); voting (Gaudet 1955); choosing an occupation (Lazarsfeld 1931); becoming a juvenile delinquent (Burt 1925); getting married or divorced (Goode 1956); joining a voluntary association (Sills 1957); moving from one home to another (Rossi 1955); and going to a psychoanalyst (Kadushin 1958; 1962; 1966). Occasionally, reasons for not acting, such as not using contraception or other family-planning procedures, are also studied (Sills 1961).

Assessment of cause

The essential difference between reason analysis and other forms of survey analysis is the method used for assessment of cause. Suppose that in a cross-sectional survey 55 per cent of those who listened to a campaign speech voted for a candidate; but of those who did not listen to him, only 45 per cent voted for him. Thus, the speech might have influenced 10 per cent of the electorate to vote for the candidate. In contrast, a reason analysis is not cross sectional, for it deals either with those who have acted or, alternatively, not acted; or, if both actors and non-actors are present (Berelson et al. 1954; Michigan 1960), it separately analyzes the reasons given by actors and nonactors. In a reason analysis, we might decide through questioning of the voters that 10 per cent voted for the candidate only because they had heard his speech (Lazarsfeld 1942). There are thus 10 per cent who may be said to have voted for the candidate “because” of the speech in both examples, but the meaning is quite different. In a cross-sectional survey the 10 per cent represents the net difference over the entire sample and is dependent on a logic similar to that of a controlled experiment, whereas in a reason analysis the 10 per cent represents the sum of the evaluations of the actions of individual voters.

Can such evaluations be valid? Any causal variate must precede or logically encompass the dependent variate (the one that it is said to “cause”) and must be linked to it through certain mechanisms. Because the respondents in cross-sectional surveys are not randomly assigned to states of the causal variate (Selvin 1957), it is possible, in principle, for the observed relationship to be a spurious one. In the example cited above, we know that persons more favorably disposed to a candidate are somewhat more likely to listen to his speeches to begin with (Berelson et al. 1954), and so the speech undoubtedly did not have a full 10 per cent effect. Theory determines whether such variates are likely to be found, and it also indicates the mechanisms, or intervening variates, that link the causal to the dependent variate. Reason analysis does not have the same statistical criteria of causal relationship that cross-sectional surveys have (Lazarsfeld & Rosenberg 1955, p. 389), and it is therefore even more dependent on theory. To say that an actor acted because of a given factor is to say that other factors were not as important. Reason analyses generally fail if they do not begin with a model of behavior that specifies all the relevant factors that might impel an action or prevent one.

When to apply reason analysis

Reason analysis can always be used in studying the subjective factors in any course of individual action. In addition, reason analysis is the method to be preferred if one or more of the following conditions hold true: a process is being studied; the act is extremely frequent or extremely infrequent; only those performing the act can conveniently be located or followed. It is therefore sometimes used for studying nonsubjective acts, such as automobile accidents (Haddon et al. 1964; Suchman 1961).

If one wants to know how an action came to be —what steps were taken and what the key choices were; what the actor thought he was doing and how he felt about it; what influences were present and what triggered the action; and, finally, what outcomes the actor expected—then no technique other than reason analysis can be used. One cannot meaningfully ask persons who did not apply to psychiatrists when they decided their problems were not severe enough to merit psychiatric attention; meaningful “why not” questions apply only to situations in which there is a social expectation of action. Moreover, even if actors and nonactors are asked their reasons, the format of reasons is rarely identical.

Frequency of action

If acts are either extremely frequent or relatively rare, comparisons with nonactors are expensive and perhaps not even meaningful, since the exceptions may be so different from the rule as to make comparison irrelevant. Almost everybody has an occupation; hence occupational choice can profitably be studied through reason analysis (Lazarsfeld 1931). On the other hand, since very few go to psychiatrists, a random sample turns up too few cases to work with. Thus, one survey (Gurin et al. 1960) found that only 42 of 2,450 Americans over 21 and living in private households had been to a psychiatrist. Since the criteria for pseudoexperimental matching (Zetter-berg 1954, pp. 145-148) of those who did go to a psychiatrist against those who did not are not clear, reason analysis is the method to be preferred in these and similar circumstances.

Convenience

Reason analysis is often performed simply because it is administratively convenient. Those who study the membership of an organization, the readers of a library, or other such populations may wish to find out why such persons joined the organization or came to use the facilities. Rather than embark on a new study of members versus nonmembers, they may use reason analysis and thus infer the reasons for joining.

The same applies to studies of long-term processes. While multiwave panel analysis is useful for such purposes and can actually be combined with reason analysis, the latter, since it is a retrospective method, is more easily suited to the study of long-term decisions.

Models of action and accounting schemes

A model of action—that is, a list of the basic elements in terms of which human action can be described, together with some notion of how action proceeds—is essential to the development of an “accounting scheme.” Such a scheme contains an organized list of all the factors that, for the specific purposes at hand, are said to produce or inhibit an action. An element of action is one in a broad category of factors that propel or repel. For example, a craving for ice cream may be a factor in an accounting scheme for studying ice cream purchases; being a six-year-old boy is an important characteristic of an actor that must be taken into account but is not in itself a propelling factor, even though a cross-sectional analysis might show high ice cream consumption among six-year-olds as compared to sixty-year-olds.

All current models of action used in reason analysis emphasize the interplay between the subjective point of view and needs of the actor, on the one hand, and the restraints, requirements, and stimulation of his environment, on the other (Hinkle 1963). In this respect these models are similar to those used by phenomenologists (Tiryakian 1965), although sociologists guided by this point of view have generally not undertaken surveys. The following are some models of action that have been used to guide the construction of accounting schemes.

Kornhauser and Lazarsfeld (1935), in reviewing early studies of buying, suggested that the elements of these actions could be partitioned into two categories: the individual—his motive and other mechanisms, such as the state of his knowledge; and the situation—in this case consisting of the product and the way it is sold, as well as various influences upon the individual.

Parsons’ theory of action offers a similar set of elements: the actor, his ends, the situation, the means and conditions in a situation, and the norms that govern the selection of means; all the elements in this model “appear from the point of view of the actor whose action is being analyzed” (1937, p. 44). In a later version Parsons and his colleagues (Parsons & Shils 1951) substituted, for ends, means, conditions, and norms, the notion of motivational and valuational orientations. The motivational orientation consists of cognitions (what an actor sees), cathexes (what emotional meaning he invests in objects), and evaluations (what weight he gives to objects). The valuational orientation sets standards for cognitions, cathexes, and evaluations. However, both of Parsons’ schemes have been used more frequently for developing typologies of actions and institutional norms than as guides to the analysis of concrete actions.

Bales’s interaction categories (1950), developed from studies of small groups, appear more relevant to actions in which other actors, seen face to face, are the only “objects”; this situation is not usually encountered in most areas where reason analysis has been applied. Parsons and Bales saw their schemes as related, however, and together developed another scheme (Parsons et al. 1953) that introduced a time dimension to Parsons’ model. Here action moves through four phases: adaptation, goal achievement, integration, and latency [seeInteraction, article oninteraction process analysis].

Since these models are only general guides to the elements that must be included in an accounting scheme, the reason analysis must specify the precise factors to be studied. For example, a reason analysis of soap buying would include exposure to advertising among its “pushes” in the “situation”; a voting study would, instead, include campaign messages; and a study of choice of psychiatrist would check on referral sources.

The selection of factors depends in part on the type of action to be studied. First, reason analysis is usually concerned with acts that involve some sort of conscious decision; habitual acts are probably not suited for any of these models. A second distinction is that between “depth” acts (i.e., ones which involve a change in self concept) and acts which are more casual. Depth acts are typically more painful for the actor and involve a longer time span: getting married, choosing an occupation, or going to a psychotherapist are all actions that involve a change in self concept. In contrast, buying groceries and even voting are more casual acts, which do not have such consequences for the individual. Finally, there are instrumental decisions or actions, such as entering business, buying stocks, and the like; these seem to involve more cognitive and more self-consciously rational elements than other types of decisions or actions.

The time perspective of an act determines how many separate decisions, acts, or phases must be studied to understand a given action, as well as the class of factors that enters at each point in the action. Thus, each action or decision is divided into appropriate elements and factors. A depth decision, such as going to a psychiatrist, may involve many steps: realizing one has a problem, talking to friends, reading books, deciding to see a professional and choosing a particular one. Casual acts, such as soap buying, might be divided into only two steps: predisposing and precipitating factors. Further, depth decisions tend to be harder to reverse than casual ones. The key time points to study in irreversible acts are obviously those that tend to limit further alternatives or that serve as prerequisites for the next act. In occupational choice, for example, the decision to attend a technical or an academic high school is a key to the range of occupations that a student may then consider.

Designing a reason analysis

Reason analysis itself is a complex act and thus has several stages. First, types of action involved in the subject to be studied are distinguished one from another; second, the act is divided into phases or separate acts, if this is necessary; third, an accounting scheme is developed for each act or phase; fourth, the accounting scheme is translated into a data-collection guide, which is typically an interview schedule; fifth, a calculus of factors must be developed so that the relative weight of different factors can be assessed. Finally, the results of this assessment are tabulated for the sample as a whole or for different segments of it.

These last two stages can only arise if the first four have been followed. Unless respondents are queried about each of the relevant factors, the multiple causation of an act is not apparent. Mixing several factors belonging to different elements and then setting their total equal to 100 per cent is a typical error. Another error is mixing several factors belonging to several elements into a single list, which is then factor analyzed. This method does not give the process of an action. The “dimensions” discovered may represent merely the elements of an action that should have been conceptualized to begin with. These and other similar errors have led responsible researchers to conclude that reasons are inadequate data for scientific inquiry. The fault lies, not in the reasons, but rather in the way they are collected and tabulated.

Interviews and questionnaires

Since an accounting scheme is the researcher’s own miniature theory of action, rather than a questionnaire or interview schedule, questions about the various elements of an action must be written as a usable “script,” fitted to the respondent’s frame of reference. Unspecified “why” questions reveal the respondent’s frame of reference. Therefore, beginning an interview with a “why” question allows the respondent to indicate what is problematic about a given situation. Afterward the original “why” question is dropped.

Most questions in the reason analysis “script” attempt to elicit answers that specify and verify the relationship between the actor’s own experience and the structure of his situation. For example, a respondent says he buys gasoline from the nearest station. Since the interviewer has an actual map of the neighborhood, probing may reveal that an even nearer station is not patronized because it does not sell a national brand—in this case, the accounting scheme succeeds in revealing an implicit norm. After the respondent has had his initial say, it is often useful to follow with a checklist, determined by the accounting scheme, to insure that aspects not immediately relevant or accessible to the respondent will be covered.

Questions on the timing of influences or precipitants help to unravel the structure of an action. These questions also allow the researcher to separate a sequence of acts into smaller-unit acts. As a rule of thumb, only those actions that a respondent can talk about as a whole, without skipping from one locus of events to another, are unit acts.

Most reason analysis interviews skip about a good deal until all dimensions are revealed through probing. Thus, experienced interviewers are usually necessary. On the other hand, if an act has been very carefully partitioned into units, elements, and factors; if pretests reveal a complete range of what respondents are likely to say; and if the respondents are sufficiently motivated to answer some open-ended questions, then it is possible to obtain from self-administered questionnaires results that are comparable to those obtained from interviews.

Strategies for assessing cause

The most difficult part of a reason analysis occurs after the data have been collected: assessment of cause and meaning must be performed. Since only actors (or only nonactors) are present in the data, cause cannot be assessed by comparing actors to nonactors and noting the differences between them. Other analytic devices must be used. Three strategies for analyzing factors directly revealed by the actors are getting the actor to do so; using the clinical judgment of the researchers; mechanically reducing a number of factors into a smaller set.

In simple actions, with limited accounting schemes, the actor himself may effectively make the assessment as to which element was most important in his action. This procedure is employed after inquiry has already revealed to both actor and researcher all the elements indicated by the accounting scheme. Empirical tests have shown that this method is surprisingly valid, much as the best way of discovering an actor’s interest in an election is not to make a complex index but merely to ask him how interested he is (Katz & Lazarsfeld 1955, p. 171; Rossi 1955, pp. 136-137). Of course, the actor’s own opinion of his most important reason for acting is a valuable datum even if it is not objectively correct.

The second technique is much like the first, except that in this case the researcher or a panel of judges performs the same evaluation (Lazarsfeld & Rosenberg 1955, pp. 401-409). In this assessment the researcher performs a task similar to that of historians who construct the “cause” of a historical event. Nagel (1961) suggested that historians implicitly invoke a series of logical strategies, both to assess causes and to rank them in order of importance. Such strategies can also be applied to reason analysis and, for that matter, to any “clinical” judgment of cause.

In the third technique, the “property space” (Barton 1955) formed by the various factors in the action is reduced in some mechanical fashion (Sills 1957; Kadushin 1958). If we knew for each voter (in a grossly oversimplified model of voting) that personal influence, campaign speeches, issues, and party loyalty were or were not factors in his choice, we could form a 16-cell property space (24). Reduction consists of combining the cells according to some rule. Cells are combined when elements or factors are combined or when, under certain conditions, the cells themselves are grouped. We might decide, in the example just given, that party loyalty would be a factor only if the other factors were not present or, that if personal influence were present, all other factors would be ignored. In contrast to the first two methods, the grounds for the reduction are clear and mechanical —an advantage in large-scale research. On the other hand, no exceptions for the intricacies of an individual case are allowed. Although reduction could, in principle, be performed before the collected cases are tabulated, almost all property spaces are reduced after tabulation, since the way cases cluster together on the various factors often determines how the cells will be collapsed. The meaning of these first three methods for assessing causes is inherent in the factors themselves. Statistical tabulation gives merely the total result of classifying each case separately.

In the following two methods the meaning of the causes can only be induced after tabulation. The fourth method combines statistical analysis with clinical assessment in order to induce the relative “impact” of the elements of an accounting scheme (Lazarsfeld 1942). Those actors who are judged to have acted mainly because of a given factor are compared with all of the actors who claimed to be exposed to the factor. If, among buyers, 20 of the 100 persons who saw an advertisement bought the product because of the advertisement, then the “impact” of the ad is 20 per cent. This can be compared to the impact of personal influence or to other factors in an accounting scheme. However, as will be shown, to close the argument completely one also needs to know how many nonbuyers saw the ad.

The fifth method is exactly like the usual survey analysis techniques and is called morphological analysis (Lazarsfeld 1959). It consists of using a factor or combination of factors to create types of actors or degrees of action. Further, in a multiple-stage action, or one with many choice points, not all actors have gone through the same stages; different paths may have led to the same outcome. All these differences can be exploited to produce analysis similar to that which would be undertaken if both actors and nonactors were present, except that the tables now concern various types of actors and their differing reasons for acting. Inferences drawn from these differences may aid the understanding of why they came to perform their actions. Morphological analysis can, of course, be combined with any of the first four methods mentioned above (Katz & Lazarsfeld 1955, p. 171). For example, reasons for action at one stage may be compared with reasons at another stage (Mills etal. 1950).

Locating the assessment

Despite its ability to locate explicit intervening variables and mechanisms in action, reason analysis does have some inherent limitations because it deals only with actors. Careful logic, theory, and substantive knowledge can, however, lead the analyst to powerful conclusions. This logical-empirical network of reasoning can be considerably enhanced by tying it at selected points to easily obtained data on nonactors, as well as actors. For example, knowledge of the rate of exposure among nonbuyers to an advertisement on television allows us to calculate the total effectiveness, as well as the impact of the advertisement (Zeisel [1947] 1957, pp. 170-173). Not only can demographic or other objective data be used, but also attitudinal and other subjective data from cross-sectional surveys. All that is needed is to ask in a reason analysis some of the same questions that have been asked of a relevant cross section.

The validity of reason analysis

If all the proper steps of a reason analysis have been followed, can the causes of an action be determined from reasons given by actors? There are two related problems here: the validity of subjective responses and the validity of causal assessment when only actors (or nonactors) are analyzed.

The paramount position of subjective materials in reason analysis may make some researchers uneasy: actors may not know the “real” reasons for their actions, and the researcher may thus be collecting a set of mere rationalizations. Further, there may be additional reasons behind the ones collected, so that the researcher becomes involved in an infinite regression.

If interviews are properly conducted, however, it is the analyst’s job to interpret the answers; as Freud (1916–1917) pointed out, whatever a person says must have some meaning. For example, in a study of reasons given by some college graduates as to why they did not choose to go on to graduate school (in the face of the expectation of 80 per cent that they would go on), it was found that lack of motivation was mentioned more often than low grades (Davis 1964). Yet a cross-sectional analysis showed that those who were not going to graduate school had, in fact, lower grades than those who chose to go on. Which finding is the “real” reason? Did not students with low grades “rationalize” them and say they were not interested? Actually, low grades do lead to low motivation, and low motivation leads to low grades, although a panel study would be necessary to determine which came first [seePanel studies].

One might say that a “real” reason is simply one that has been built into an accounting scheme. There may be several levels of reasons: for example, people may go to a psychiatrist “because” of an emotional problem that developed at age two or “because” of the influence of a social circle of “friends and supporters of psychotherapy” (Kadushin 1962; 1966). The focus would depend on the purpose of the study.

Finally, there is a good deal of evidence that when the findings of cross-sectional analysis are compared with those of reason analysis, they are found to be similar (Lazarsfeld & Rosenberg 1955, pp. 404-419; Rossi 1955; Burt 1925). Considerable work remains to be done, however, in enhancing and testing the validity of reason analysis, as well as in testing and improving various models of action.

Charles Kadushin

[See alsoEvaluation research; Experimental design, article onquasi-experimental design; Interviewing, article onsocial research; Survey analysis, article onmethods of survey analysis.]

BIBLIOGRAPHY

Many reason analyses are in unpublished commercial research. The best published introductions to the topic are Lazarsfeld & Rosenberg 1955; Lazarsfeld 1959; Zeisel 1947. Especially relevant in the theory of action tradition are Parsons 1937; 1947; 1960; Parsons & Shils 1951; Tollman 1951; Weber 1904–1917. Early but important methodological works in the empirical tradition are Lazarsfeld 1935; 1942. Some empirical works explicitly using accounting schemes and reason analyses are Goode 1956; Kadushin 1958; Katz & Lazarsfeld 1955; Lazarsfeld 1931; Mills et al. 1950; Rossi 1955; Sills 1957.

Bales, Robert F. 1950 Interaction Process Analysis: A Method for the Study of Small Groups, Reading, Mass.: Addison-Wesley.

Barton, Allen H. 1955 The Concept of Property Space in Social Research. Pages 40-54 in Paul F. Lazarsfeld and Morris Rosenberg (editors), The Language of Social Research. Glencoe, 111.: Free Press.

Berelson, Bernard; Lazarsfeld, Paul F.; and Mcphee, William N. 1954 Voting: A Study of Opinion Formation in a Presidential Campaign. Univ. of Chicago Press.

Burt, Cyril L. (1925) 1944 The Young Delinquent. 4th ed. Univ. of London Press.

Davis, James A. 1964 Great Aspirations: The Graduate School Plans of America’s College Seniors. Chicago: Aldine.

Fisher, R. A. (1935)1960 The Design of Experiments. 7th ed. New York: Hafner.

Freud, Sigmund (1916–1917) 1952 A General Introduction to Psychoanalysis. Authorized English translation of the rev. ed. by Joan Riviere. Garden City, N.Y.: Doubleday. → First published as Vorlesungen zur Einführung in die Psychoanalyse. See especially the second lecture.

Gaudet, Hazel 1955 A Model for Assessing Changes in Voting Intention. Pages 428-438 in Paul F. Lazarsfeld and Morris Rosenberg (editors), The Language of Social Research. Glencoe, Ill.: Free Press.

Goode, William J. 1956 After Divorce. Glencoe, Ill.: Free Press.

Gurin, Gerald; Veroff, Joseph; and Feld, Sheila 1960 Americans View Their Mental Health: A Nationwide Interview Survey. Joint Commission on Mental Illness and Health, Monograph Series, No. 4. New York: Basic Books.

Haddon, William; Suchman, Edward A.; and Klein, David 1964 Accident Research: Methods and Approaches. New York: Harper.

Hinkle, Roscoe C. 1963 Antecedents of the Action Orientation in American Sociology Before 1935. American Sociological Review 28:705-715.

Kadushin, Charles 1958 Individual Decisions to Undertake Psychotherapy. Administrative Science Quarterly 3:379-411.

Kadushin, Charles 1962 Social Distance Between Client and Professional. American Journal of Sociology 68:517-531.

Kadushin, Charles 1966 The Friends and Supporters of Psychotherapy: On Social Circles in Urban Life. American Sociological Review 31:786-802.

Katz, Elihu; and Lazarsfeld, Paul F. 1955 Personal Influence: The Part Played by People in the Flow of Mass Communications. Glencoe, 111.: Free Press. -” A paperback edition was published in 1964.

Kornhauser, Arthur; and Lazarsfeld, Paul F. (1935) 1955 The Analysis of Consumer Actions. Pages 392-404 in Paul F. Lazarsfeld and Morris Rosenberg (editors), The Language of Social Research. Glencoe, 111.: Free Press. → First published by the American Management Association as “The Techniques of Market Research From the Standpoint of a Psychologist.”

Lazarsfeld, Paul F. (editor) 1931 Jugend und Beruf. Jena (Germany): Fischer.

Lazarsfeld, Paul F. (1935) 1954 The Art of Asking Why. Pages 675-686 in Society for the Psychological Study of Social Issues, Public Opinion and Propaganda. Edited by Daniel Katz et al. New York: Dryden. → First published in Volume 1 of the National Marketing Review.

Lazarsfeld, Paul F. 1942 The Statistical Analysis of Reasons as Research Operation. Sociometry 5:29-47.

Lazarsfeld, Paul F. 1959 Reflections on Business. American Journal of Sociology 65:1-31.

Lazarsfeld, Paul F.; and Rosenberg, Morris (editors) 1955 The Language of Social Research: A Reader in the Methodology of Social Research. Glencoe, 111.: Free Press. → See especially section 5.

Lipset, Seymour M.; Trow, Martin A.; and Coleman, James S. 1956 Union Democracy: The Internal Politics of the International Typographical Union. Glencoe, 111.: Free Press. → A paperback edition was published in 1962 by Doubleday.

Michigan, University Of, Survey Research Center 1960 The American Voter, by Angus Campbell et al. New York: Wiley.

Mills, C. Wright; Senior, Clarence; and Goldsen, Rose K. 1950 The Puerto Rican Journey: New York’s Newest Migrants. New York: Harper.

Nagel, Ernest 1961 The Structure of Science: Problems in the Logic of Scientific Explanation. New York: Harcourt. → See especially pages 582-588.

Parsons, Talcott (1937)1949 The Structure of Social Action: A Study in Social Theory With Special Reference to a Group of Recent European Writers. Glencoe, 111.: Free Press.

Parsons, Talcott 1947 Introduction. In Max Weber, The Theory of Social and Economic Organization. Translated and edited by A. M. Henderson and Talcott Parsons. New York: Free Press.

Parsons, Talcott 1960 Pattern Variables Revisited: A Response to Robert Dubin. American Sociological Review 25:467-483.

Parsons, Talcott; Bales, R. F.; and Shils, E. A. 1953 Working Papers in the Theory of Action. Glencoe, 111.: Free Press.

Parsons, Talcott; and Shils, Edward (editors) 1951 Toward a General Theory of Action. Cambridge, Mass.: Harvard Univ. Press. → A paperback edition was published in 1962 by Harper. See especially Chapter 2.

Rossi, Peter H. 1955 Why Families Move: A Study in the Social Psychology of Urban Residential Mobility. Glencoe, Ill.: Free Press.

Selvin, Hanan C. 1957 A Critique of Tests of Significance in Survey Research. American Sociological Review 22:519-527.

Sills, David L. 1957 The Volunteers: Means and Ends in a National Organization. Glencoe, 111.: Free Press.

Sills, David L. 1960 A Sociologist Looks at Motivation. Pages 70-93 in Nathan E. Cohen (editor), The Citizen Volunteer: His Responsibility, Role, and Opportunity in Modern Society. New York: Harper.

Sills, David L. 1961 On the Art of Asking “Why Not?”: Some Problems and Procedures in Studying Acceptance of Family Planning. Pages 26-36 in All India Conference on Family Planning, Fourth, 1961, Report of the Proceedings: 29th JanuarySrd February 1961, Hyderabad. Bombay: Family Planning Association of India.

Suchman, Edward A. 1961 A Conceptual Analysis of the Accident Phenomenon. Social Problems 8:241-253.

Tiryakian, Edward A. 1965 Existential Phenomenology and the Sociological Tradition. American Sociological Review 30:674-688.

Tolman, Edward C. 1951 A Psychological Model. Pages 279-361 in Talcott Parsons and Edward Shils (editors), Toward a General Theory of Action. Glencoe, Ill.: Free Press.

Weber, Max (1904–1917) 1949 Max Weber on the Methodology of the Social Sciences. Translated and edited by Edward Shils and H. A. Finch. Glencoe, Ill.: Free Press.

Zeisel, Hans (1947) 1957 Say It With Figures. 4th ed., rev. New York: Harper. → See especially Chapters 6-7.

Zetterberg, Hans L. (1954) 1965 On Theory and Verification in Sociology. Totowa, N.J.: Bedminster Press.