The ideal of exact science—to arrive at laws and predictions by way of experiment and measurement—has been adopted by many social scientists. Whether they investigate minority problems, group attitudes toward some topic, or communication contents, they invariably formulate hypotheses enabling them to break down their material into quantifiable elements. Applied social research is largely a matter of coding, scoring, counting and tabulating, with much thought given to statistical technicalities and the reliability of the procedures. Since social phenomena differ from the subject matter of exact science in that they are historical entities and as such carriers of unique values and qualities, we may also define the current major goal of our social sciences as follows: they aim at transforming quality into quantity, at revealing unchanging nature behind ever-changing history.
Much can be said in favor of this approach. The continuum of social phenomena ranges from cases of highly individual behavior to cases of uniform mass behavior, and whenever we concentrate on the latter we move through a region ruled by the law of averages, a region in which many problems of interest lend themselves readily to quantification and statistical elaboration. And there seems to be no reason why the same procedures should not be applied as well to the analysis of intensive interviews, propaganda communications and other more individual phenomena at the opposite end of the conditions. They too include characteristics that may profitably be measured and computed; it all depends on the questions we pose, on the answers we expect. Quantitative analysis is a legitimate attempt to establish a body of strictly objective knowledge in fields where theory has for a long time indulged in unverified speculations and apparently uncontrollable, if perhaps cogent, descriptions. It is inevitable that the exponents of a science aspiring to the status of exact science should reject wholesale what they vaguely call the “impressionist” approach. They indict it for relying on untouchable “intuition” rather than rational probing, on appropriate evaluations rather than estimates in clear-cut figures.
All this is justified in a way. However, there is sufficient evidence that the course our social sciences are taking, leads them ever farther away from basic problems and relevant questions. In their eagerness for statistical accuracy they lose sight of the driving powers behind our social life, leaving it to the historians or political scientists to survey expanses in which they themselves are seized with agoraphobia. Their preoccupation with ever-refined breakdowns makes them disregard or even overlook anything that cannot neatly be pigeonholed. Sheer technique threatens to become an end in itself. This shows distinctly in the treatment of social phenomena conspicuous for their individual features and approached with a view to bringing them out. It is the kind of job assigned to content analysis. Take a sample of VOICE OF AMERICA broadcasts to some foreign country: what matters much is their structure; for instance, do they capitalize on the effects that may be produced through an appropriate linkage of the successive bits of the news they convey? Such effects obviously deepen the impact of the news items themselves. But because of their involved nature these structural qualities of the material resist the mechanics of coding and, hence, cannot be processed in content-analysis fashion. Only qualitative analysis, however, decried for its alleged “impressionism,” might be able to capture them. In other words, content analysis is even in this relatively simple case not in a position to isolate, and account for, vitally important content. Its seeming objectivity is bought at a price; its very goal prompts it to strip social entities of their individual fullness. It is certainly no coincidence that current social research deals largely with problems arising from managerial concerns. How are we to handle groups or masses for this or that purpose, in this or that situation? In tackling these problems on the level of collective behavior, quantitative analysis is in its element, for it need not inquire into the situation or the purpose itself, nor appraise any of the qualitative factors which emerge on more individual levels. Quantifying techniques conform perfectly to managerial techniques.
In short, our social sciences are in a critical condition. Their insistence on emulating exact science threatens them with starvation. Yet the only alternative left to them would seem to be their surrender to the very impressionism of which they are trying to rid themselves. Is there a way out of this dilemma? The time has come, it appears, to take a good look at the foundations of the social sciences. We therefore propose a research project reconsidering the conceptions currently held of quantitative and qualitative analysis. This project should be based on the following two hypotheses.
First Hypothesis: the breakdown of individual wholes into quantifiable elements involves a series of subjective considerations which inevitably color the ensuing measurements, qualifying their objectivity. To verify this hypothesis it seems advisable to examine the customary procedures of content analysis because of its special concern with highly individual material. We submit that, in processing much material, the content analyst is time and again bound to make assumptions and decisions which cannot deny their “impressionist” origins. He may happily count the number of times the word “America” occurs in an editorial, but in doing so he presumably neglects various indirect references to the concept “America”; evidently, his predilection for the computable must be traced to the “impressionist” assumption that it is more relevant to the meaning of the text than those uncomputable references. Or he may decide to score an attitude under the heading “very favorable” instead of “favorable,” but he would be blind to qualities indeed were he not, in so deciding, haunted by some of the scruples inseparable from the “impressionist” approach. Even the tiniest element challenges the analyst to review the infinite contexts to which it belongs. Each waterdrop contains a universe.
This is rarely acknowledged, though. Content analysts are not accustomed to think in terms of qualities, and they follow rules and directives which farther prevent them from going astray. Yet despite these protective habits and conventions which, of course, make for high reliability, “impressionist” evaluations and preferences actually persist all down the line. The objectivity of applied social research and in particular content analysis imitates rather than equals that of exact science. It is always an exhilarating spectacle to watch rows of tidy figures, complete with decimals, floating like little islands in an ocean of doubts and pretending to unattainable exactness. This is said not so much to deprecate quantifying techniques as to point to their limits. They are extremely valuable when applied to social phenomena sufficiently primitive and widespread to admit of quantification without a noticeable loss of quality; they fail to yield results when applied to highly individual textures.
Second Hypothesis: Qualitative analysis—a convenient designation of hermeneutic techniques in these fields—is more reliable, more scientific, than the champions of content analysis are inclined to believe. But having made a bogey of the “impressionist” approach, they naturally throw the baby out with the bath and refuse to acknowledge the legitimacy of any breakdown that cannot be presented in tables and charts. It is unfortunate that under the impact of their preferences, increased by the aversion of most exegetes to concern themselves with quantifications, qualitative methods have rarely been applied and never systematically cultivated. There is a palpable lack of pertinent examples: they are found in scattered books, papers and reports, and till now little attention has been paid to their far-reaching implications. In the interest of our hypothesis we will have to study these few existing models. For the moment all that can be said in its support are the following considerations. Exactly like content analysis, qualitative analysis proceeds to break down a given whole into small elements or units; thus enabling us to disentangle the fabric of its qualities and compare them with those other, similar wholes. But unlike content analysis, it is not primarily interested in matters of quantification. Some of the elements at which it arrives may be quantifiable and therefore should by all means be coded, scored and computed; others, as has been exemplified above, obstruct any attempt at numerical formulation. What counts alone in qualitative analysis is the selection and rational organization of such small elements as are expressive of the essentials of the whole.
But how can we find out about these essentials without collapsing into plain subjectivism? The danger is not so great as it appears. For even though such subjective factors as the analyst’s philosophical viewpoints, political convictions, etc. unavoidably influence his conception of the whole he is analyzing, they will to a large extent be naturalized in the process. Whether or not he states them overtly from the outset—he should indeed—they are bound to leak out anyway. And once the cards are on the table, these influences and their distorting effects can in a measure be controlled and discounted. For the rest, access to the essentials of social entities depends upon familiarity with anything that might have contributed to their emergence, a sense of history, a flair for ideological currents, and some experience of human behavior—not to forget the greatest virtue of all: circumspection in weighing and balancing against each other the various data assembled.
We submit that these abilities can be taught to the degree to which they are needed for the purpose at hand, and that it is equally possible to train students in the successive operations that lead from the first rough scheme of analysis to its satisfactory completion. The skills involved in carrying out mental experiments or imagining possible alternatives are as teachable as are the skills of a flyer. No “intuition” of the kind dreaded by content analysts is required to trace and make evident the structural characteristic of VOICE OF AMERICA broadcasts, despite their allergy to statistical procedures; nor is it likely that different qualitative analysts will account for them in different ways.
In fact, we contend that on many occasions the evaluations of qualitative analysis exceed the measurements of quantitative analysis in reliability and precision. The reason is that the latter, in its desire to reach the haven of statistics, often simplifies or obscures aspects of the material which are a genuine concern of the former. But how can we ascertain the validity of qualitative findings? Qualitative analysis insists on making evident any of its results; it aims at inclusive objectivity, the very objectivity which is aspired to in the humanities. To be sure, it does not, and can not, provide the evidences of exact science; yet this kind of evidence is unattainable to content analysis also.
At this early stage a detailed blueprint would seem premature. We prefer to offer, for further discussion, a few scant hints concerning the execution of the planned project and its specific aims.
The first step, we suggest, is to assemble and analyze existing examples of qualitative analysis, with a view to codifying the methods used in them. In a way it is a preparatory step.
We suggest that research itself be conducted in close connection with a large-scale project involving diverse quantifying techniques, especially content analysis—a project of the kind carried out by Columbia’s Bureau of Applied Social Research and similar institutions.
From beginning to end an intimate working community should be established between the staffs of the two projects, thus affiliated, so that the researchers engaged in qualitative pursuits are in a position continually to compare notes with those in charge of quantifications.
Finally we suggest a staff which, if possible, does not exclusively consist of social science students. It might be a good idea to include a historian and perhaps a psychologist.
Provided research is organized in this way, we shall be able
1) to examine, step by step, quantifying procedures from the angle of qualitative analysis and, perhaps, to define their inevitable limitations;
(Some pertinent problems: Is there a point beyond
which the measurements of content analysis turn
out to be meaningless? What portion of the
material lends itself to statistical elaboration
and what portion does not? Etc.)
2) to undertake a qualitative analysis of each parts of the material as demonstrably elude the grasp of the affiliated project;
3) to train the staff, prior to and during this undertaking, in qualitative techniques for the purpose of testing out contention that they are teachable;
4) to study, in collaboration with the staff of the affiliated project, the methodological and practical implication of this whole experiment—in particular the problem of how qualitative and quantitative analysis might complement each other in future social research.