When selecting the research method it is usually advisable to consider whether you can base your work on an earlier theoretical model. Sometimes a model, even a preliminary one, can help your work decisively, and in such a case it will also affect the logical process of analysis. There are three alternatives which are discussed in more detail later on:
Research is exploratory when you use no earlier model as a basis of your study. The most usual reason for using this approach is that you have no other choice. Normally you would like to take an earlier theory as a support, but there perhaps is none, or all available models come from wrong contexts.
On the other hand, even when there is relevant theory and models, sometimes you may prefer not to use them. Reasons for this can be:
Exploratory research means that hardly anything is known about the matter at the outset of the project. You then have to begin with a rather vague impression of what you should study, and it is also impossible to make a detailed work plan in advance.
The gradual process of accumulating intelligence about the object of study means also that it will be impossible to start by defining the concepts of study. You have to start with a preliminary notion of your object of study, and of its context. During the exploratory research project, these provisional concepts then gradually gain precision.
In the absence of tried models and definite concepts you must start the exploratory study from what you have: one or more objects of study. It is common that in the beginning of exploratory study you will take a holistic look at the objects. It means that you start by gathering as much information about the objects as possible, and postpone the task of cutting away unnecessary data until you get a better picture about what is necessary.
Any object can be looked at from several different viewpoints, either from the angles of various established sciences or just from miscellaneous practical points of view. As soon as possible, you should specify the viewpoint of your study and explain how you understand or "take" the object. This does not mean that you have to to start your work by clarifying the essence of your object of study, i.e. what the object really is. Instead, you should try to contemplate and clarify how you see the object: should it be defined on micro level as a result of the individuals' instincts, drives and experiences, or maybe on macro level as an expression of development in society.
The method of alternating point of view (like in the diagram above) can even be used as a research method. It is especially suited to an explorative researcher working alone. It will deepen his understanding and can sometimes reveal valuable new aspects to the topic, cf. Hermeneutic Research.
The progress of a project of study becomes easier as soon as you have defined your point of view and your problem. After this, you will need to gather only such empirical knowledge that is related to the problem; that will enable you to restrict the material you will have to analyse. This does not mean that you should disregard all the cases that do not fit into your conjectures - sometimes anomalies or surprising cases can point the way to important amendments or corrections to existing theory.
Sometimes it is difficult to define what is relevant in advance; it only becomes apparent through analysis. In such a case you can simply start by studying one single specimen or case which illustrates the interesting problem, and then you continue studying a gradually growing number of objects until it becomes apparent that you cannot get deeper into the problem. An indication of such a "saturated" state of study is that the study of new items or cases no longer reveals new interesting information. You will often need to gather quite a lot of material before you can define the final goal of your project, and a large part of this material will not be used in the final analysis.
The exploratory analysis of empirical field observations starts by checking that the field reports are written down intelligibly and without ambiguity. Often the original reports have been made in hurry; in that case they should be clarified by the initial observer or interviewer. The same person is often best adapted to extricate the significant findings from observations because he/she is able to judge which details are important and which can be left out. In the same time he/she can start building a preliminary model from those patterns which seem to recur often, or estimate how well an earlier known model fits the observations.
As soon as the invariance in the data becomes apparent you can omit all the material that is no longer relevant and compress the remaining, relevant information. This compacting is usually done with the help of coding the typical and frequent elements, that is by assigning short names, letters or other symbols to them. By cross-tabulating the symbols you can get an overall view of all the material, and it will be easier to uncover its structure or rearrange it so that a latent structure becomes visible.
Analysis in exploratory research is essentially abstraction and generalization. Abstraction means that you translate the empirical observations, measurements etc. into concepts; generalization means arranging the material so that it disengages from single persons, occurrences etc. and focuses on those structures (invariances) that are common to all or most of the cases.
It will seldom be possible to divide exploratory study into such clear phases as is common in the case that the object has been studied earlier. According to Alasuutari (1993 p.22), in qualitative analysis of empirical findings, you can distinguish two phases but these two overlap:
In the simplification phase, the material is inspected from the theoretical point of view of the study project, and only the points relevant from this angle are noted. Details differing from one individual to another at random are omitted or pushed aside so that the general lines of the data can be discerned more easily.
Simplification continues by finding the relationships between separate observations or cases. Some tools for this work are comparison and classification. The goal is to find the general rule or model that is valid in all or most of the observations. This model can be, for example, development or evolution, causality, or a conscious action to attain an outcome which is typical in normative research. -- In any case the analysis starts from separate cases and aspires to create one or a few general models.
"Solving the enigma" does not always mean answering exactly those questions that were asked at the outset of the project. Sometimes the most interesting questions are found at the end of the research, when the researcher has become an expert on the subject. It is often said that "data teach the researcher".
The purpose of descriptive exploratory research is to extract a structure from the source material which in the best case can be formed as a rule that governs all the observations and is not known earlier (per the definition of exploratory study). Finding the unknown structure may need some creative innovation, because even the most sophisticated computerized analysis methods cannot automatically uncover which type of structure is concealed in data. Usually you first have to formulate a tentative pattern for the assumed structure in the observations and then you can ask the computer to estimate how well the data corresponds to the model, cf. Tools for Analysis.
In normative studies the exploratory approach is unusual, because the normative target - improving something in the object - in general engages with a known theoretical background which you can take as a basis of your study, thus shifting to the usually more effective method of Research on the Basis of Earlier Theory which is explained in the next paragraph.
Nevertheless, sometimes it happens even in normative study that the direction of desirable improvement is initially unclear, and your only choice is to start with the exploratory approach. Such is the situation when you know that the present state of the object of study is unsatisfactory but you do not know exactly what is wrong in it, neither do you know of any superior usable substitutes for it. For example, in the initial stage of action research everybody perhaps agrees that the present mode of working is unbearable but all known remedies seem inapplicable, and the participants therefore start making from empty table a descriptive model of the work to be used as a basis of development.
Many of the problems of exploratory research can be avoided if the researcher can start with a model, developed in earlier studies, which he uses as a "working hypothesis". The model can either consist of cases (holistic model) or of concepts (analytic model). During the analysis, the researcher tries to see whether the collected material conforms to the model or must he correct the model or look for a more suitable one.
Often the study simply proceeds by enlarging an earlier model. A good rule to be followed in such a situation is: Start from what is known. Proceed by enlarging the mapped area, and connect the new intelligence to the known facts. Sometimes all that you need is only an adjustment of a few details in the existing model. This is often the case when the study shall give grounds for a forecast or new product development and the environment of intended application is slightly different from the one of the earlier study.
The existence of a tentative model helps in selecting the logical structure of the entire research project and planning it. The model helps you to decide which material has to be collected, from which cases or specimens and about which attributes or variables of these cases. Even the recording of observations is facilitated because often you will be able to utilize earlier definitions of variables. The same applies to analysis methods: often you can borrow them from earlier works.
In descriptive study the project is often arranged as distinct phases, like in the diagram above. First you demarcate the population about which you need knowledge, then select a sample, gather the empirical data, analyze them, perhaps with the same method as in the earlier study from which the model was taken, and finally assess the findings.
Adopting models from earlier treatises involves a risk: it can affect your observations so that you wrongly discard the anomalies or those cases which too much differ from what would be expected on the basis of the old theory. If this happens, you will never discover the weaknesses of the old model.
In normative study models are used for describing the existing problems and defining the improvements to the object of study. If you can find an existing descriptive model of the object, made in an earlier study, you can often transform it into a normative model by adding an evaluative dimension to it. For example, the model of industrial production on the right can be made normative by adding the dimension of profitability, and a target for it.
Methods for analyzing information and evaluations with normative models are discussed in Normative Case Study , Normative Comparison , Normative Classification , Normative Study of Variables and Normative History.
Once the target for development has been defined with the help of a normative model, the project often continues as planning the practical operations, perhaps also realizing them and measuring the results. Sometimes the same model can be used as a basis of all these operations, like in the figure on the left, but usually you will have to refine a model successively several times in the process of transforming a definition of goals into a plan of action or into a design of a product. The latter process, for example, can include such phases as product concept, various drafts of design, a series of prototypes and finally a detailed proposal for the product.
Optimally a normative research project proceeds through successive stages:
It is quite usual that you will have to repeat the above sequence several times before you get an acceptable result. Normative projects often deal with complex practical problems, and when making a theoretical model of the problem, the researcher may wish to make the model more easily manageable by simplifying it, i.e. by leaving out factors that seem nonessential. However, in the final practical test or appraisal it may turn out that an excluded factor is important after all, which makes it necessary to adjust the model and repeat the sequence once more.
Sometimes the object of study is already well known and you just want to investigate its behaviour in a specific situation. In such a situation you can choose to construct a hypothesis, i.e. an expectation of the behaviour of the object, or a preliminary answer to the question that you are studying. You are usually free to decide if you want to use one or not. E.g., if you want to learn if x really equals two times y, you can set as your hypothesis
x = 2y
During your project you then collect empirical data which allow you to test your hypothesis and see if it is true or not.
Hypotheses are always based on analytic models, and they are often causal. They are always accurately stated and quite often stated as an arithmetic model, like for example
y = f(x)
x = the independent variable,
y = the dependent variable.
The above hypothesis includes only one variable of each type; there are, however, usually more of them in real research projects.
If you choose to use a hypothesis, you should plan the logic around it in the way that Bunge (1967, 9) explains:
It is seldom - perhaps never - possible to reach an absolute certitude when verifying a hypothesis. This is the case especially when the hypothesis is intended to hold true anywhere, i.e. also for the cases that are similar to those that have been examined. Therefore most modern researchers accept in practice the idea that when speaking of 'truth' of a hypothesis they actually mean verisimilitude or credibility. This distinction, nevertheless, has no decisive consequences in practice: you can use 'credible' findings exactly in the same way as 'true' findings.
Disturbances. Usually the object of study is influenced by various factors besides the independent variable mentioned in the hypothesis. These disturbances, sometimes called "noise", prevent the researcher from clearly seeing the influence of the independent variable. Such factors whose systematic influence is known beforehand can simply be eliminated by making a suitable correction in the measurements. Unknown factors which cause detrimental random variation in the dependent variable are more difficult to handle. The researcher can be prepared for disturbances and for the random variation of the explained object in alternative ways:
The method of hypothesis was originally developed for descriptive studies. This type of research aspires to get factual knowledge about the object of study, and the criterion which is used in accepting or rejecting a descriptive hypothesis is factuality or truthfulness. The same criterion is applied to other phases of descriptive project as well, as is explained in Assessing Input Data , Assessing Correctness of Analysis and Assessing Theoretical Output.
In normative research a hypothesis is seldom used as such, but it is interesting to note that a normative project often includes a decisive focal point which determines the project's success or failure in much the same way than the test of a hypothesis does in a descriptive project. This focal point is the final practical testing of the normative proposal. Because the target in normative study is not just to get information but primarily to improve the object of study or other similar objects, the principal criterion when testing normative development proposals is not truth but instead practical value and functional operativeness. This criterion is discussed in Evaluating Normative Proposals.
Another difference to testing a descriptive hypothesis is that the process of development need not end in the test. If the first proposal must be rejected the normal practice is to prepare and test another proposal. This means that the process of development returns to that of normative research on the basis of earlier model, discussed above.
Tests of normative proposals are habitually carried out when developing a new product. Their procedures are discussed in Presenting the Draft and Prototype and Evaluating a Design Proposal. Similarly, in the development of an activity the project usually includes practical testing of the proposals, as described in Assessing Activity Development.
August 3, 2007.
Comments to the author:
Original location: http://www2.uiah.fi/projects/metodi