At numerous points in this manual the suggestion has been made to the operator to "analyze the content" of this or that material. For the most part, the suggestion has related in printed materials, such as newspapers, magazines, books, codes of laws, lists of organizations, lists of names of office holders. At some points, it has also been suggested that the operator analyze personalized material, which would be written down but not necessarily printed, such as diaries or other personal documents, answers of informants to questions, replies of individuals to questionnaires. In addition, there has been some mention of non-printed content, such as that of speeches and radio programs. All the communication forms mentioned have been subjected to successful content analysis along restricted lines.
Thus, a definition of content is suggested; it is simply any body of symbols. By symbol is meant any sign or word or number or other means of conveying information. For the most part, the operator will be dealing with content which is made up of words, although he may have content with a high and useful percentage of number symbols. This approach can be extended to think of a picture as being a body of content made up of a number of symbols, or of a piece of music satisfying the definition. Successful content analyses have shown trends in styles, motives, and scope of paintings and musical compositions.
The size of the "body of content" can vary tremendously; it may be a single one-hour speech, or it may be 100 years of the files of a newspaper. As far as the operator is concerned, the body of content will always be limited to that which is pertinent to his needs. Thus he may be interested in only a portion of a list of names or organizations, or only the editorial pages of a newspaper, or a series of radio programs which lasted for two weeks. In terms of sampling methodology suggested in Appendix A, the "body of content" becomes the "universe" or "population" which he is studying. The universe of content, then, might be all of the editorial pages of X newspaper published in a two-year period, and the actual "sample" studied might be a randomly selected 25 pages drawn from the 730 making up the universe. Such sampling is very frequently done in content analysis to reduce the bulk of the material. Within the limits of sampling error, it is a very useful kind of reduction.
Whenever the suggestion to analyze the content has been made in the manual, several basic assumptions have been made at the same time. The first is that the content is not a random collection of symbols, but that it is a patterned collection of symbols which convey information, i.e. the content is a communication. The second is that the pattern is cumulative, i.e. there will be more of some kinds of symbols than of others, and that the frequency of these symbols will tell something about the content. Third, it is assumed that the symbols can be quantified, or in plain words, may be counted in some systematic fashion. Fourth, it is assumed that by counting the symbols and analyzing the resulting tabulations, one can discover more precise and more useful information about what the body of content contains than by some other method.
Almost all bodies of content satisfy the first two points. Not all satisfy the third, and many fail to meet the fourth. Only when all of these assumptions can be satisfied should the operator engage in content analysis. In following pages means of testing content in advance to see how it measures up to the "criteria" will be given.
LEVELS OF CONTENT ANALYSIS
While specific details of "content analysis" have not been mentioned in the manual, it is apparent that there are several levels of activity which have been thought of when this term has been used. Three levels may be identified : (a) subjective; (b) mechanical ; (c) subjective-mechanical.
By "subjective analysis" is meant simply the reading or listening to or looking at any body of symbols and endeavoring to abstract from the multitude of symbols which it contains one or several central messages. This is the natural habit of any reader or listener. One spends two days reading a book; to an inquirer, he describes in a few words that the book is about. Or again, one listens to a speech for an hour and when a friend asks what the speaker said, one answers briefly with a summary of the information as he sees it. It might take the form: "Jones spent the whole hour talking about the danger of communism."
That there are serious limitations to this approach to what any body of symbols contains is readily apparent. One may misread the book, or one may fail to cumulate the points which might have been cumulated, or one may over-emphasize certain content because it fits his own point of view. Similar problems may limit the effectiveness of subjective analysis of a spoken body of symbols. There may be additional problems of simply remembering what the speaker said at the beginning and what he said at the end; or there may be problems of not hearing clearly, all of which may affect the reliability of the analysis. On the other hand, there are many advantages of this subjective kind of study. It is quick and effective. Actually, no one works with symbols without using it.
While the operator will often be forced by the exigencies of time to be satisfied with such an analysis, it should be noted that it is not content analysis in a scientific sense. One of the experts on the subject, Bernard Berelson, says that "content analysis is a research technique for the objective, systematic, and quantitative description of the manifest content of communication." Reading over a communication will give one information on its "manifest content," but it is difficult to assert that such a method is objective, systematic or quantitative.
To get away from the lack of rigor and the ever-present problem of bias of the subjective approach to content, one can make what has been termed here a "mechanical" analysis of the material, which would meet the demands of the above definition. To return to the example of the speech, such as an analysis might result in changing the subjective statement: "Jones spent the whole hour talking about the danger of communism" into "Jones mentioned the danger of communism seventeen times in his one-hour speech." If there were several speeches in which the operator were interested, such a mechanical counting might make it possible to move from the subjective statement: "It seems to me that Jones speaks more about the danger of communism than Smith," into the more precise statement, "In his last speech Jones referred to the danger of communism seventeen times while smith in his speech only mentioned it three times."
Such a mechanical approach may provide a relatively small amount of additional information over the subjective analysis of a generalized book or speech. But when one is confronted with materials which has more pronounced repetitive characteristics, the mechanical analysis may be very much more useful than the subjective approach. Thus, if one were analyzing a copy of Whos Who subjectively, he might flip through the pages and discover that there seems to be a concentration of men from certain eastern colleges and universities; or he might note that there seemed to be quite a few Episcopalians, or not as many Catholics as he expected to find. These subjective judgments about the contents of Whos Who would be rather weak, however, unless the analyzer were unusually proficient in counting things in his head as he went along. And again, his special interests, his own biases, might make it possible for him to be quite sure that certain characteristics were predominant, when in fact they were not.
If the mechanical approach were taken instead, the analyzer would record the mentions of various universities and possibly the presence or absence of college education, the religious background, the political party membership and several other dimensions suggested by the biographical material. And when he had finished, he would have a rather precise analysis of the content of the Whos Who. He could now say that 43 per cent of the men mentioned came from twelve eastern universities, that 25 per cent were Episcopalians and that 22 per cent were Catholics (or whatever the figures showed). This would provide him with much more useful information about the content than the subjective estimates mentioned earlier.
Since a relatively low percentage of the available collections of symbols are organized in the highly repetitive form of a Whos Who however become necessary to use a more complicated system of content analysis. It is dubbed here the "subjective-mechanical" system, since it involves (at least for the operator) a considerable measure of subjective choice, yet applies much of the rigor of the simple counting method. Such a procedure might be more effective in telling what Mr. Smith and Mr. Jones were talking about. For example, it could have been that Jones, despite his seventeen mentions of the "danger of communism" actually put all but one of them in contexts which showed that he was de-emphasizing the "danger" while Smith, with only three mentions, put all of his into contexts which emphasized them. Thus, a better approach than the simple mechanical would be not only to count the mentions of the phrase, but also to evaluate each use of it as "emphasizing" or as "de-emphasizing."
Specialists in content analysis endeavor to define rigorously what they will regard as "emphasizing" or "de-emphasizing" or "strong-weak", or "favorable-unfavorable" in advance of beginning such an analysis, so that several different individuals will be able to code the content in the same fashion. They have gone a long way toward removing the subjective nature of such evaluations. On the other hand, the operator, working under the conditions that he must, had best remember that there are subjective factors when he departs from mechanical recording of symbols.
PROBLEMS OF CONTENT ANALYSIS
It is undoubtedly apparent that there will be many problems in the use of this scientific tool. They may be thought of conveniently as those of the pre-counting period, those of the counting period, and those of the analysis of data period. Actually, most of them relate to the first period. This is particularly so far the operator. Because of his lack of time and personnel, he must work out in advance almost all of the problems which may be anticipated, otherwise he may find that he has put in many hours of work and has failed to get a useful result.
1. Knowledge of the body of content. A first pre-requisite to useful content analysis is to have a reasonably good knowledge of the body of content one proposed to study. This must be obtained by the subjective kind of analysis which was set up as the "first level of operation." It may be possible for the research scientist to start counting symbols on the suspicion that his totals may reveal something of interest, but it is not wise even for him. The pre-knowledge should include general information on what the material contains, something about the kinds of symbols which are present, something about the frequency of those symbols, some pretty good hunches on what patterns exist and what they may tell if they are measured more precisely.
This pre-knowledge of the content has many additional facets. If one is dealing with newspapers and hopes to make a time study over several months or years, he had better make sure that copies of the paper will be available in the pattern than he chooses. If he hopes to compare the output of several publications, all must be available and contain comparable materials. If he is dealing with newspapers or magazines, he should probably have some information on their political affiliation or pre-disposition. If he is dealing with a directory, he ought to know whether it represents a comprehensive coverage of the persons in the field it essays to cover, or whether it is merely a "vanity book" in which entries represent the payment of fees. He ought to know whether the 1955 edition represents a complete revision or just a bringing up to date of material gathered in 1939.
In summary, he ought to know enough about the material to predict with some accuracy the nature of the pattern he expects to find, the classifications he will need to record symbols referring to the pattern, the rough proportion of quantity of frequencies.
2. Language Problems. Eventhough the content may be in a foreign language, the operator may find it possible to make useful content analyses of the mechanical type, provided that he has a minimum knowledge of the language. It would be possible for a person with a very modest knowledge of a language to make a fairly good analysis of a Whos Who or other biographical compendium, provided he learned the limited vocabulary used, the system of abbreviations, and other characteristics of the particular work.
But in order to make an accurate and meaningful analysis of a general body of content in a foreign language, a considerable expertness is required, particularly when one moves in the direction of the subjective-mechanical approach. Any attempt to measure the "direction" of content along a "favorable-unfavorable" or "strong-weak" continuum requires ability to catch the nuances of a language, to understand its idiom. Even a simple word count approach may offer some problems. One group of students who measured the occurrence of a set of symbols in newspapers of five different nations (in four languages) noted: "No symbol list is perfectly translatable between languages." In order to make their lists comparable, they found it necessary to translate some key words into several possible synonyms in the foreign language, and at other times found it necessary to limit the use of a particular word to contexts which provided the meaning they wanted.
3. Hypothecation. To make content analysis useful, the operator must be prepared to make a "hypothesis," to develop a theory, or to have a strong hunch about what his analysis will show. It might be stated in this form: "If I count the frequencies of this list of symbols, it will show that the elite are predominantly of X background," or any variant of this which is pertinent. What is important is to have a theory of what the analysis will show before engaging in it. To return to the Whos Who example, it would be eminently foolish to start recording the universities attended by the persons listed, unless one could predict in advance that the elite were clustered among a certain small group of schools (and that knowledge of this fact would be of aid to the operator). In terms of elite, analysis it will always be necessary for the operator to have an expectation that the content analysis will tell him something about the elite in a more precise fashion than he presently knows. Further, he must decide in advance whether this information will be useful to him.
It can be said that almost any content analysis will provide more precise information about the material under study. But sometimes the information amounts merely to "making the obvious precise." The operator cannot afford the luxury of using this time-consuming tool for such a result.
While the need of a working hypothesis about the result in advance of analysis has been stressed, the operator should consider once he has decided to use it the possibility that the gathering of some small additional amount of information may provide him with more insight into the patterns of the content. Thus he might "risk" a small amount of additional time on the hope that information will "emerge". While the possibility of "emergent hypothesis" should not be overlooked it should not be the main motivation.
4. Categorization. The construction of a hypothesis about the content will usually suggest categorizations. Several levels of categorization will be required. They will vary with each individual problem, but they will generally fall into one of the four levels to be listed here: (1) They will involve the setting up of classes of symbols which be recorded or counted. Such would be the key words relating to the hypothesized pattern, the presence or absence of a particular school or religious faith, etc. (2) They will involve setting up classes of referents of the recording symbols, i.e. the persons, ideas, nations, things to which the recording symbols refer. (3) They will involve setting up classes or sub-divisions of the total body of content, as articles, or pages, or columns, or items. (4) They may also involve setting up classes of context which will show the direction of the symbol in relation to the referent among some continuum, as "strong-weak" or "favorable-unfavorable."
Suppose the material were the ubiquitous "Whos Who and the hypothesis that there was a predominance of persons form a small group of universities. The major class of recording symbol will be schools of higher education; to reduce tabulation, it may be decided to set up geographical classifications or to provide evidence on the hypothesis, a list of ten or twelve schools might be made recording symbols, with all others to be placed in miscellaneous. The problem of referent is also simple each recording symbol will refer to one person listed in Whos Who. The organization of the volume suggests a ready made subdivision of the total body of content, the single biography. If the study were expanded to record religion and political party, the list of recording symbols would be expanded, but referents and content subdivisions would remain the same.
Suppose the hypothesis were that the laws of the nation provided more special privileges for the clergy than for any other elite group. Here it would be necessary to list classes of symbols which would relate to "privilege." Advance knowledge of the general tone of the law and subjectively obtained information on the content of the total code would perhaps suggest that recording symbols be classified broadly into "economic privilege," "privileges before courts," "enforcement of deference to ...," etc. The actual symbol list might be of groups of words as presented in sentences, or paragraphs of the code, rather than in individual words. The referents would be elite classes. If the law sometimes granted privileges and sometimes withheld them, it might also be worth while setting up categories of context, such as "grants indulgence" or "withholds indulgence." The organization of the legal code might suggest the paragraph as the content subdivision unit, or the section, or some other unit inherent in the organization of the material.
5. Special Problems of Referents. Most of the time, the operator will automatically have the elite (or potentially elite) individual or organization as referent. Yet there will be cases which call for another referent. If he is investigating attitudes of the elite as exemplified in their personal documents or in their special publications or in their answers to questions, the referents may become national states or governments or political parties or ideological concepts.
The more concrete the referent the simpler will be the design of the analysis. Vague, inclusive referents make it difficult to tie the recording units down. Content analysis has been described as determining "Who is saying WHAT about WHOM with WHAT EFFECT." The "about whom" which is more precise will be more useful. Thus the use of particular individuals as referents, or groups of individuals who can be identified quite positively, will make the analysis more meaningful.
6. Special Problems of Context Units. In most cases, the operator should probably avoid the problem of context by so designing his study that useful material can be obtained without measuring the context within which a symbol is found. However, it may be imperative to measure the direction of a body of content, and in such case, the checking of context cannot be escaped. Thus one might have a hypothesis that the press of country X which formerly was hostile to the U.S. has now become more favorable. This immediately raises the problem of how big an area of context is necessary to determine whether the use of a symbol is favorable or unfavorable. One approach might be to record the mentions of U.S. (and synonyms for it) and to record them as "favorable-neutral-unfavorable" on the basis of the individual sentence in which they occurred. The sentence, "The U.S. has the finest air force in the world," provides a mention which one has no difficulty as describing as favorable. But suppose the next sentence says: "However, the basic power in military conflict today is not air power but land power." The juxtaposition immediately suggests the use of a larger contextual area. But this solution is not without pitfalls, for as the sign of context units is increased, the judging of direction becomes more and more subjective.
Despite this limitation, it can be said that the speed (and possibly the accuracy) of analysis is improved by the use of larger contextual units. If the operator is in doubt on a particular piece of content, he may test by scoring direction on the basis of a smaller unit, and then on the basis of the larger unit. In most cases, the two approaches will provide similar scores, and the larger unit has much economy as far as time is concerned, as well as in reducing the number of ambiguous findings of direction.
7. Choice of Content Subdivision Unit. In cutting up the body of content into subdivisions, it is also advantageous to have them as large as possible. But an even more useful approach is to have the subdivisions correspond to the natural subdivisions of the body of content. Several examples which have been given have suggested just that. In Whos Who it was the individual biography; in a newspaper it might be the item; in a code of laws the section or paragraph. One of the difficulties of the natural unit is that it may vary in size considerably. This may not be important as far as the biographical item is concerned, but it may be quite important in recording information about newspaper articles. In the latter case, while the item is preserved as a unit, it may be scored in a measure which will reflect variations in size, such as the column inch. Recording in this fashion, however, does not remove all of the difficulties of equating units exactly. Very slight variations in type size may mean considerable variations in the gross content of an equal number of column inches; the same can be said for slight variations in column widths, which are hardly discernible to the naked eye. These factors become important when one wishes to compare the gross content on a particular subject, for example, in one newspaper with that in several others.
If time is not as much of a factor, smaller units may be used. One could go to the "word" as unit of content, subdividing the material, in effect, into individual words and recording the presence or absence of chosen recording symbols. A somewhat larger unit of content (also used as a contextual unit or recording unit) is the phrase, also referred to as the "theme" or "statement." This requires breaking down sentences into the several statements they may contain. The sentence, "The most powerful organization in this city is The Bankers Club," breaks down into "The Bankers Club is an organization in this city," and "The Bankers Club is the most powerful organization in this city." This approach makes necessary the sub-division of more than half of the sentences of English content, since complex ideas are produced with about that frequency. Such an approach to German, because of the nature of the language, would produce a much higher ratio of statement for sentences, and would increase the amount of time necessary for the analysis.
In analyzing spoken communications a time-unit is sometimes chosen. Thus radio programs might be analyzed by 15-minute segments, or a speech by five-minute segments.
This discussion of units is by no means definitive. The nature of the content and of the design of the analysis may suggest others which will be pertinent and effective.
8. Choice of recording Unit. Similar variations in size are possible for recording units as for the several other types which have been discussed. At the outset, it should be noted that the analysis will be simpler, if the recording unit is made as small as possible, in direct contrast to the recommendations made on the other units. For as the size of the recording unit is decreased, it becomes more precise; the word is easier to record accurately than the "theme," the theme than the paragraph or item, etc.
Even in the selection of the word as a recording unit, however, problems may arise. Is the analyst to record only the occurrence of the specific word, or shall he accept synonyms, slang expressions, figurative expressions? Thus, the decision to record the symbol "U.S." or "United States," raises the problem of also recording "America" (in the sense of the U.S.), "Uncle Sam," "The Land of the Free," "the leading power of the western states," "The outstanding imperialist nation in the world," etc. The analyst should explore such possibilities in advance of recording, and he can expect that even with careful consideration additional synonyms may require consideration. Or he may decide, on basis of knowledge of the content, that he will record only the concrete symbol and ignore synonyms.
9. Sampling Problems. Detailed information on sampling has already been presented in Appendix A, but a few special problems of content sampling will be touched upon here. One relates to the sampling of any printed material issued serially, such as a newspaper, magazine or other periodical. Such issuance usually means that there is a pattern of presentation of information which tends to make certain issues of the week, month, or quarter specialized in function. As a result, sampling must be designed to prevent an under or over-representation of these differentiated issues.
The daily newspaper in America, to illustrate this point, has a concentration of religious content in its Saturday issues. Thus, if one were seeking a representative sample of daily newspapers, the lack of Saturday copies would mean that the religious news would be missed while any concentration of Saturday copies beyond 1/7 of the sample total would over-represent the amount of religious news. In the case of a daily publication with such a pattern, the choice of an interval such as 5 or 15 will automatically prevent the "serial" quality of the publication from over-representing a particular day of the week.
This characteristic of publications usually is related to a basic serial pattern in the human events which the publications describe. Thus a sample overweighted with October issues of newspapers or magazines in the U.S. would over-report news of elections, while LAN overweighing of April issues of a Russian publication would over-report information concerning preparations for May Day celebrations. This characteristic of both events and publications can be used by the operator to his advantage when he seeks information about a particular class of events. Here he can concentrate his study in the time period most likely to carry the information he seeks.
Some mention of sampling levels also should be made. In any study of printed materials, the first level of sampling would be of titles thus one might take a sample of five from all of the newspapers of a region, or a sample of three magazines out of a class containing 80. The next level is that of issues within the selected tittles thus one might take every fifteenth issue of a newspaper issued during five years. A final level of sampling would involve taking a certain part of each issue, such as every other page, or only the front page, or some other part of the total.
If the sampling at the several levels is random in the sense explained in Appendix A, it can be said that the final sample represents the total body of content. On the other hand, the subjective choice of a "sample" at any particular level, would make it necessary to reduce the total universe about which one spoke when he reported his final results.
10. Coding Problems. Moving to the period of actual recording of data, new problems are discovered, although they are not independent of those already discussed. Coding problems, which refer to difficulties arising when one endeavor to record the occurrence or the direction of the various symbols which have been decided upon, are actually separate from the choice of unit problems only in a time sense. The greatest danger here lies in the possibility that coding may not be consistent, that identical uses of symbols may be recorded as favorable at one time and as unfavorable at another, or that synonyms for the chosen symbol may be counted at one point and not another, or that subdivision units may not be held constant.
In a large scale content analysis project many coders may be employed, and the problem is augmented. Different coders may have different understandings of units, or of the decided-upon definitions of context which will be considered favorable or unfavorable. Many tests have been made of coder reliability (the degree to hitch two or more individuals code the same material in the same way.) The results almost never show perfect reliability, but in well-planned operations using competent and well-instructed coders correlations of .9 are frequently achieved. (A perfect correlation would be 1.0 and an absolute lack of correlation would be .0).
If the operator hopes to turn coding over to assistants he will have to face this problem. A rigorous training period in advance is necessary to obtain coder reliability as high as .9. If the operator has to depend upon personnel of relatively low skill, he may find it necessary to restrict the analysis to the recording of simple information so that there will be a minimum of opportunity for coder interpretations which may produce variations in recording.
If the operator does his own content analysis, the problem is reduced. Nevertheless, if he sets up a complicated scheme with finely shaded classifications and all-inclusive units, he may discover that his own reliability from one part of the content to another is not very good. A simple test of reliability is to take a complicated piece of content which has been scored sometime previously and to restore it. Here, however, there will be a tendency to remember the previous scoring which will obscure a possibility to "drift" away from earlier practices in coding a long and complicated body of material. A better procedure would be to request another individual to code a body of content which the operator has coded, and to note differences.
If the single coder will keep his criteria in mind, however, he can code reliably. And what is especially important, he will know the kinds of decisions that went into the production of this or that total score for a symbol or context. Further, a single coder can adjust a pre-arranged unit or coding system to accept an unexpected kind of symbol or context better than two or more coders can, since he does not run the risk of having the unexpected situation produce two interpretations which may be inconsistent.
11. Emergent Hypothesis. Some mention of the possibility of "Emergent" hypothesis was made in Paragraph 3. As the operator is actually coding, he may discover signs of a pattern which suggests a new hypothesis, perhaps unrelated to the original one which he set up. Such a result is one of the potential "bonus" values of content analysis and should be utilized whenever possible. It might require a refinement of the coding system to reveal it in greater clarity. A change, of course, may require returning to the previous material and rechecking.
13. Analysis: Statistical Significance: When frequencies have been cumulated and the whole array of data is available, the operator must turn to analyzing what he has obtained. One of the first factors is that of statistical significance. Here, he will apply the suggestions of Appendix A, at least roughly, to discover what degree chance might have played in his results. Actually, this phase might be little more than recognition that he CANNOT apply statistical tests of significance, because of the method of selecting his sample.
14. Analysis Comparison with outside criteria. One of the best tools of analysis of data will be comparison with outside criteria of one form or another. If the operator has analyzed content which will show social class of the elite, he will put his obtained information along side of information on class distribution of the total population. If he has obtained information on education, the same kind of comparison is suggested. If he has information on other elites, or on elites of the United States, there may be the possibility of making cross-cultural comparisons.
One of the problems of symbol counts, seeking to determine the direction of content of a publication, arises here. The operator may be without any criteria. His result of 47 per cent favorable might be quite precise, but it will mean little unless he can put it in relationship to some other figures. If such criteria are lacking, the operator may have to break down his material into time periods, so that he can show change at least within the body of content studied; or it may be possible to relate the changes in content to outside events, or to data which has been obtained through interviewing informants.
15. Analysis: By Inference. Sometimes it will be possible for the operator to set up a series of logical steps from the result of his content analysis to an estimate of elite behavior or intentions. One of the most interesting cases of this type came during the war, when British content analysis correctly inferred from studying the content of Nazi propaganda, that the Germans had a new secret weapon, the V-2. The opportunity to analyze the propaganda output of an elite, or of a particular member of an elite, will be available to the operator and often he may find it impossible to make inferences on the basis of such analysis.
Heavy emphasis should be given to the necessity of applying all other factors known about the elite in making such inferences. While content analysis is a useful research tool, it is not a magic one. Another problem of inferences based on analysis of propaganda content, is that sometimes the elite behavior will be the inverse of what the face indications of the analysis show. In the V-2 case, the inference was based upon positive references (in advance of the firing of the rocket bombs) to secret weapons, plus the fact that this class of propaganda had never been used in the past in the "Cry Wolf" sense. However, another case of inference by the same kind of approach, (which involved correct estimates of the Nazi elite perception of their Italian allys will to fight) involved a reverse twist. Here the propaganda content showed a tendency on the part of the Germans to assert positively that the Italians were staunch allies. The correct inference was that the Nazi elite thought they were not.
RELATING THEORY TO PRACTICE
It is recognized that in many cases the operator will find it impossible to reach even the simpler levels of content analysis which have been suggested here. Nevertheless, the operator will find it possible to use even subjective analysis more effectively, if he has become aware of the possibility of more rigorous methods. For one thing, he may detach some of the positiveness which he has attached in the past to his subjective evaluations of content. For another, he may be more aware of the patterns and cumulative characteristics of the content, so that his subjective analysis will be sharper, more likely to catch the indications of the pattern.
If he finds it possible to use even a modest program of content analysis over a period of time, he will build bases for comparisons, so that his occasional "soundings" will provide him not only with a look as of a given moment, but also with information about changes in the composition of the elite.
This is not to say that content analysis, even in its most complicated sense, will become a magic thermometer from which he can take periodic readings on the temper of the elite, or on its composition, but it will provide him with more precise information than he has available by guessing and simply playing hunches.