The Holy Grail: In Pursuit of the Dissertation Proposal

Michael Watts, Institute of International Studies, UC Berkeley

To Page 1 of this essay

Page 4 of 5

Entry Points and Using Evidence

At this point let me step back a little and reflect upon how we identify a researchable problem or question (what I shall call points of entry), and the ways in which such a question or problem can be framed (what I shall refer to as logics of inquiry). Often we start will an ill-defined interest that takes the form of an association or a broad relationship, ill specified and general in its articulation. For example, we might be interested in the relation between migration and intra-household dynamics, or between Hindu nationalism and "neoliberal" reforms, or between armed struggle and forms of democratization. Quite how we get to these entry points and why really does not matter -- and we should not spend too much time figuring out why we are drawn to violence or gender or class conflict (though these might be interesting topics for you and your therapist). These are all important entry points -- and like all entry points they leave out important sorts of middle level questions and specifics: what forms of armed struggle; what are the specific aspects of neo-liberalism and how do they have causal efficacy, what sort of evidence would we need and use to identify this or that variable.

Entry points then usually take the form of a particular sort of question or query, with the goal naturally to identify the "right" research question. Often this process is treated as one of individual choice or by a curious process of osmosis in which the field of knowledge is transmitted to the researcher, or that it emerges inexorably from the data. In practice there is of course a complex tacking back and forth between theory, question and data. One cannot over emphasize the importance of struggling to formulate a coherent -- that is to say conceptually integrated and empirically grounded -- research question. The question does ultimately commit or obligate the scholar in keys ways: to mastering literatures, to identifying with a theory, of plowing through sources of data and so on. All of this is likely to lead to dead ends and paralysis unless the researcher is explicit and self-conscious about the theoretical and empirical decisions one has made.

Whatever the entry point, you will need at some point to generate a specific question rooted in empirical circumstances and with a particular design and scale (perhaps a large n, perhaps a national comparison, perhaps a single village case). An entry point typically generates different sorts of questions, each if which may provide the groundwork for the elaboration of a research program. One sort of question -- practical -- might emerge for a student's experience working in a non-profit or a government agency. How can an Indian NGO better delivery family planning advice to south Indian women in deeply patriarchal male dominated households? How might organic grape growers in Napa Valley improve their market share? My experience is that students who have strong political commitments to their research and who have returned to graduate school from say practical work on development projects in the Third World, often lean toward such action questions. They may be driven say by the frustrations of western aid projects to target particular communities or by the tensions between local NGOs and their transnational partners. But such concerns must be located with respect to a theoretical framework, and within a logic of inquiry, if they are to be action-research (that is to say a theorized and scholarly program of work with direct practical implications emerging from the object of study). Another entry point and research question is empirical. Empirical questions can also take a variety of forms: some are abstract ("how is class consciousness shaped by social interactions among persons of equal status"), some are concrete ("were Muslims less involved in the genocidal activities in Rwanda in 1991 than Catholics"), or historical ("how did the language of the 1946 strike in X differ from the same plant's strike in 1978"). And finally some questions are theoretical: "does bureaucratic domination reduce the legitimacy of rule"? "Under what historical circumstances does social integration increase or decrease"? "How do members of militant movements construct beliefs about the meaning of life which justify suicidal acts?"

The question then becomes, how do I push this question forward, develop and refine it, convert a hunch into a research program, a proposal. There are several immediate sorts of responses to this impulse. One is to figure out a conceptual toolkit that can help you refine your question, but can also generate hypotheses or propositions to be tested or evaluated. Another is the sort of evidence that is appropriate to the questions and the means by which valid evidence can be collected. A third, is how a particular approach to linking evidence and theory is shaped by practical considerations: your limited time, energy and resources. In quality research institutions much time is rightly spent on theory and on the student having acquired a road map of theory appropriate to the discipline, and also appropriate to the selection of concepts that are relevant to the research project. Much less attention is often given to the perhaps banal and pedestrian questions of evidence: both what constitutes evidence for a particular approach to a problem (and why), and the mundane issues of acquiring such evidence however constituted. As I have already mentioned, it is customarily the "methods" section of the research proposal that is weakest. It is often weak because it is underspecified -- "I shall engage in participant observation"- but also because the connections -- I would say the rules -- by which evidence is linked to theory or theorized claims is often opaque and unclear.

Let's take three projects for illustrative purposes. One is a study of a farmer's movement in India with a focus on the question of the meanings of being identified with the movement. Another examines the particular historical conjuncture out of which the Mafia was born in mid nineteenth century Sicily. A third is an analysis of strike action in relation to rational choices made by differing sorts of actors. One way to see these different sorts of questions -- they might all incidentally be approached in Marxist or Weberian terms -- is that they fall into one of three logics of inquiry: respectively, they are phenomenological, historical and causal. (I have taken the following discussion from an excellent but alas unfinished and unpublished book project by Professors Bob Alford and Paul Lubeck on Social Science Theory and Method that emerged from their teaching research design to sociology students at UC Santa Cruz during the 1980s. For similar discussions you may wish to see Andrew Sayer, Method in Social Science, London, Hutchinson, 1992.) Such logics provide ways for linking theory and evidence and also help you see the sorts of choices you have to make regardless of the content of the question. The logic of inquiry does not help you answer your questions; it highlights the choices that have to be made (working in one way with one set of tool does provides limits of what can be pursued and how) and their consequences. Logics of inquiry offer you a way of formulating and reformulating your question within different approaches, and to see the choices available to you.

Let me examine each of these logics in turn as a way of showing how something about the rules linking theory and evidence, and differing logics confer differing choices and options.

Causal Logic: one broad class of procedures attempts to distinguish the relative importance of different causal factors, to discover the causal structure that explains variation in the social world. It explains variations in the attributes of different units of analysis by deploying a multivariate analysis. In order for evidence to be recognized by theory (whether Marxian, rational-choice or Foucauldian), it must be transformed into "variables". This approach is frequently grounded in and draws strength from positivism (the model is of course the natural sciences, the world is assumed to be knowable and real, observations can be replicated, bias controlled and the world is divisible into autonomous parts). The most important variables cannot be manipulated by the investigator who must assume that classification into subgroups substitutes for experimental manipulation. It is assumed that one can draw data from a sample and measure the variables of interest without rupturing the actual social relations among individuals and groups from which the data is drawn. A survey is the most typical quantitative example of multivariate analysis. To work, some degree of independence of the independent variables must be assumed and defined. Objectivity is the careful specification of variable and their measures, and the reporting of all relevant data and how they were gathered. The observer is assumed to be at some distance from the observed. A basic task is obviously to reduce interview bias and measurement error. A model of causal logic might be Durkheim's study of suicide.

Phenomenological Logic: This is an interpretive logic of inquiry. The various theories that make use of it assume that social reality is constructed by and through symbolic and cultural interpretations, webs of meaning and signification built and used by human actors. It is typically based upon a phenomenological philosophy and is customarily associated with field observations of real life situations, participant observation, ethnographic method and secondarily the interpretation of key texts. Within this logic there is a sort of causal connection between categories in the actor's mind and their actions; between the roles being played and the rules of the game. But as Lubeck and Alford say, the open-ended negotiated, self-conscious character of social interaction means that causation is not linear; relations are contingent and subject to continual change. Meaning symbols and discourses are the theoretical categories that identify and locate relevant evidence for analysis. Observations of actual interactions, events, movements and gestures would be the typical qualitative data. Participant observation is the method that links phenomenology to interpretive theory and to qualitative field notes as the form of evidence. Objectivity results from self-conscious checking of the observer's perceptions and his relations to those observed. The researcher participates in social life and categories of observation cannot be separated from those activities. While such questions of meaning -- for example which symbols are struggled over in political struggles of X -- are associated with cultural theory, and the humanities, but there is no a priori reason why surveys might endeavor to collect systematic data on some symbolic questions. A model of interpretive logic might be Weber's Protestant Ethic.

Historical-Dialectical: This is approached by be based on a historicist philosophy, and draws strength from the observation and belief that contingent sequences of events take place within an interdependent historical totality. Evidence is primarily textual and the method is to construct a narrative sensitive to conjunctures, contingencies and contradictions. Historical analysis assumes that all relationships and processes are interdependent and change over time in relation to one another. The essential concepts are totality (a single case changing over time), conjunctures (overdetermination and multiple factors changing together), and chronology (sequences of concrete events). Historical events are discrete moments in time that can stand in for a variety of forces at work within a totality. Theoretical categories that identify empirical units of observation are, for example, the Depression, the Great War, and the New Deal. They sum up the meaning of a particular period and each of these events is a complex totality which derives its meaning from a larger context but also becomes the mechanism for gathering and interpreting specific historical data. As Lubeck and Alford say "the ideal type example of the historical logic of inquiry focuses on a single case seen as a totality of interdependent elements which constitute each other and cannot be separate from their relations from each other. The sequences of events are contingent outcomes which cannot be attributed to separable causes". One might say this inseparability is dialectical. A search for patterns and changes is the method linking philosophy of history to historical theory, and the unit of analysis is the global, societal or sub-societal entity that has constitutes a whole. The interplay between structural forces and conjunctural or contingent events is an intrinsic theoretical issue within the historical logic of inquiry. There is a sort of causation at work here too but causes are neither linear nor independent; they are interdependent and dialectical. A model might be Marx's Brumaire.

These logics are abbreviated and stylized of course but I want to refer to two key points about them. First, each type of evidence for a project located with respect to one of these logics must be converted to the appropriate form recognized by the theory in order to be defined as appropriate for explanation. A causal theory only recognizes primary data that can be converted into a variable. Texts or narratives of events are key to historical logics but must be converted into variables through some sort of coding if they are to be deployed by causal logic, although this coding may be qualitative as well as quantitative. Interpretive theory may use field notes but within the historical logic they are a text and for causal analysis they must be rendered into multivariate form. Second, in practice a research project may deploy two or more of such logics of inquiry -- great works typically do -- and a research program may indeed be involved in using specific data in a variety of ways (if possible) to make it appropriate for different types of analysis. Whether and how for example a historical text can be converted into a variable is an important and complex question. The point I seek to emphasize however is that analyses of quite different sorts located in different theoretical traditions may all locate their study in one of these logics. Marxist, neoclassical and institutional analyses of household economic behavior may all adopt a sort of causal analysis by deploying similar sorts of multivariate data. Similarly a Marxist analysis could be located in theory in any of the logics of inquiry (though I appreciate there will be a ferocious debate over whether causal logics are consistent with some versions of Marxian political economy). The key point however is that focussing on these differing logics makes clear to you the sorts of choices that are available to you once a question has been formulated.

Once you have made your choices -- your Marxian analysis of the culture of work in south Indian textile factories -- you can begin to seriously explore the sorts of evidences you need and the knotty questions of validity, reliability and so on. This is not the place to work through such a complex field but I would in passing take note of a number of issues that are typically lost sight of in many of the sorts in international fieldwork-oriented projects that pass over my desk:

• National Accounts: virtually all dissertations addressing some aspect of development typically refer to and make use of macro-economic and national accounts data (even if the object of scrutiny is the village or the household). Yet anyone who has worked in Africa or Indonesia is acutely aware of deep problems associated with the most basic economic data (for a period in the 1980s for example the Nigeria Central Bank published no financial and monetary data; the disparities between World Bank, FAO and USDA estimates of say Senegalese food output can be enormous). All of which is to say the epistemology of numbers warrants more attention than is customarily granted to the duplicated World Bank table or the UNDP statistical roll.
• The Archive: the use of colonial archives has also become an almost standard part of foreign area field research (and the same can be said of many other historical sources -- Missionary archives, business archives and so on -- that are deployed by the social sciences). I raise this point because rarely is the question addressed in a research proposal: how can you be confident that you can derive the sorts of data you need from historical texts? This question is not only one of textual interpretation, but also of whether such information was indeed collected and whether and how it can be located! Just because you are interested in prostitution in colonial Nairobi or communal violence in colonial South India, does not mean that the archive itself (and its organization) is laid out in a fashion which will expedite the discovery, or indeed the interpretation, of the information you need. To simply invoke the archive as a source of evidence then is simply a beginning, not an end. As Luise White discovered in her book on prostitution in Kenya one needs in some way to understand the social and epistemological organization of the archive -- the "colonial mind" -- in order to figure out where certain sources of information might be located.
• The Assistant: Even though many dissertation projects have quite limited budgets, the use of assistants (for surveys, as translators) is commonplace. Much has been made in Anthropology of course of the deployment of the "informant" or assistant. I simply want to raise here the practical dimensions of using enumerators and assistants. How in other words one recruits (from where, with what background, with what local understanding and connection) assistants, how they are to be trained, their contractual or other relation to you the Principal Investigator, their salaries and benefits; in other words the dull details of employment, and the hermeneutic complexities of a sort of intellectual Intermediation (you are getting information twice removed). Whether all of this needs to be documented in a research proposal is an open question. But once again to simply indicate in a Methods section that you will make use of 'interviewers' can only raise flags unless this is framed in some way.
• The Survey: Much could be said about surveys and this is not the place. In lieu of a full discussion, I wish to make the following points. Survey design is an art in it self and any project involving large n samples and a survey designed by the Principle Investigation (PI) must establish that they (the PI) have the training to undertake such a project. Here the absence of such courses on many campuses is striking and the utility of summer courses at some place like the ISPCR at the University of Michigan is accordingly magnified. Second, surveys generate substantial amounts of data, and a proposal must therefore be able to address the demands and resources associated with large scale data collection, management and analysis (saying that you have put in the budget the $5000 request for a new powerful laptop will not do it!). And third, the survey (however constituted) is something that some sections of the social sciences and the humanities shy away from ("I do not collect that sort of data", "I prefer ethnography" and so on). In keeping with the thrust of my remarks and the value of multiple methods in research design, I would encourage students to think about surveys in a variety of way, not least the fact that a survey even if it is not a central data collection device is a powerful tool for scanning, probing and assessing the landscape on which your study will be located. In other words, there can be spillover effects and insights derived from the collection of a rather mundane baseline survey. It has also been my experience that the need for systematic data -- which can only be generated by a survey -- may emerge in the course of a project that did not anticipate the need for such data. Being prepared for such eventualities then has a particular payoff.

Next page: Warnings, Pathologies, and Conclusions

Process & Parameters Pages

1. Introduction
2. The Funding Regime: Selection Criteria and Processes
3. Primary Objectives and Parameters
4. Entry Points and Using Evidence
5. Warnings, Pathologies, and Conclusions