Most modes of observation require the researcher to intrude to some degree on whatever he or she studying. However, Durkheim’s analysis of suicide did nothing to affect suicide. His study is an example of unobtrusive research, or methods of studying social behavior without affecting it. Unobtrusive measures can be qualitative and quantitative.

In content analysis, researches examine a class of social artifacts that usually are written documents such as newspaper editorials.

Historical/comparative analysis is a form of research with a history in the social sciences. The main resources for analysis in historical research are historical records.

In 1966, Eugene Webb and three colleagues published an ingenious little book on social research (revised in 1981) that has become a classic. It focuses on the idea of unobtrusive or non-reactive research.

Do you want to know what exhibits are the most popular at a museum?

You could conduct a poll, but people might tell you what they thought you wanted to hear or what might make them look intellectual and serous. You could stand by different exhibits and count the viewers that came by, but people might come over to see what you were doing. Webb and his colleges suggest you check the wear and tear on the floor of various exhibits. Those who have the most worn tiles are probably the most popular.

The possibilities are limitless. Like a detective investigating a crime, the social researcher looks for clues. If you stop to notice, you will find that clues of social behavior are all around you. In a sense, everything you see represents the answer to some important social scientific question – all you have to do is think of the question.

Content analysis is the study of recorded human communications. Among the forms suitable for study are books, magazines, Web pages, poems, newspapers, songs, paintings, speeches, letters, email messages, bulletin boar posting on the Internet, laws, and constitutions, as well as any components or collections thereof. Shulamit Reinharz points out that feminist researchers have used content analysis to study "children’s books, fairy tales, billboards, feminist nonfiction and fiction books, children’s art work, fashion, postcards, newspaper rhetoric, clinical records, research publications, introductory sociology textbooks, and citations, to mention only a few."

Content analysis is particularly well suited to the study of communications and answering the classic question of communication research:" Who says what, to whom, why, how, and with what effect?" Are popular French novels more concerned with love than in the United States? Was the popular British music of the 1960s more politically cynical than the popular German music during that period? Do political candidates who primarily address "bread and butter" issues get elected more often than those who address issues of high principle? Each of these questions addresses a social scientific research topic. Although such topics might be studied by observing individual people, content analysis provides another approach.

An early example of content analysis is the work of Ida B. Wells. In 1881, Wells, whose parents were slaves, wanted to test the widely held assumptions that black men were being lynched in the South primarily for raping white women. As research method, she examined 728 lynching reported during the previous ten years. In only a third of the cases were the lynching victims even accused of rape, much less proven guilty. Primarily, they were charged with being insolent, not staying in their place.

Some topics are more appropriately addressed by content analysis than by any other method of inquiry. Suppose that you have interested in violence on television. Maybe you suspect that the manufactures of men’s products are more likely to sponsor violent TV shows than are other kinds of sponsors. Content analysis would be the best way to find it out. Briefly, here’s what you do. First, you would develop operational definition of the two key variables in your inquiry men’s products and violence. Ultimately, you need a plan that would allow you to watch TV< classify sponsors, and rate the degree of violence on particular shows. Next, you would have to decide what to watch. Probably you would decide what stations to watch, for what period, and at what hours. Then, you would stock up beer and potato chips and start watching, classifying, and recording. Once you have completed your observations, you would be able to analyze the data you collected and determine whether men’s product manufactures sponsored more blood than did other sponsors.

As a mode of observation, content analysis requires a thoughtful handling of the "what" that is being communicated. The analysis of data collected in this mode, as in others, addresses the "why" and "with what effect."

Take the example you cannot watch TV all days and nights long by yourself. Usually, it is appropriate to sample.

Units of analysis are the individual units that we make descriptive and explanatory statements about. For example, if we wish to compute the average family income, the individual family income, the individual family is the unit of analysis.

It is essential that this issue be clear, because sample selection depends largely on what the unit of analysis is. If individual writers are the units of analysis, the sample of the writers are the units of analysis, the sample design should select all or sample of the writers appropriate to the research question. If books are the units of analysis, we should select a sample of books, regardless, of their authors. Bruce Berg points out that even if you plan to analyze some body of textual material, the units of analysis might be words, themes, characters, paragraphs, items (a book or letter), concepts, semantics, or combinations of these.

In content analysis, we could employ any kind of the conventional sampling techniques. We might select a random or systematic sample of novelists, laws, and etc, stratified sampling of newspaper (group all newspapers by regions, size of community, language, frequency of publications, and circulation), cluster sampling, etc (politician represents a cluster of political speeches).

Two ways of coding - coding the manifest content or coding latent contest.

Manifest content is the visible, surface content. To determine, for example, how erotic certain novels are, you might simply count the number of times love appears in each novel or average number of appearance per page. Or, you might use a list of words, such as love, kiss, hug, or caress, each of which might serve as an indicator of the erotic nature of the novel. This method would have the advantage of ease and reliability in coding. It would have a disadvantage in terms of validity. Surely the phrase erotic novel conveys a richer and deeper meaning than the number of times the word love is used.

Latent content is underlying meaning of the communication. In the present example, you might read an entire novel or a sample paragraph or pages and make an overall assessment of how erotic the novel was. Although your total assessment might very well be influenced by the appearance of words such as love and kiss, it would not depend fully on their frequency.

Clearly, this second method seems designed for tapping the underlying meaning of communications, but its advantage comes at a cost to reliability and specificity. Especially if more than one person is coding the novel, somewhat different definitions and standards may be employed. A passage that one coder regards as erotic may not seem erotic to another. Even if you do all of the coding yourself, there is no guarantee that your definitions and standards will remain constant throughout the enterprise.

For all research methods, conceptualization and operationalization typically involve the interaction of theoretical concerns and empirical observations.

The categories researchers use in a content analysis can be determined inductively and deductively. If you are testing theoretical propositions, your theories should suggest empirical indicators of concepts. If you begin with specific empirical observations, you should attempt to derive general principles relating to them and then apply those principles to the other empirical observations.

If you plan to evaluate your content analysis quantitatively, your coding must be amenable to data processing. This means, first, that the end product of your coding must be numerical.

Second, your record keeping must clearly distinguish between units of analysis and units of observation, especially if they are different. If novelists are your units of analysis, and you wish to characterize them through the content analysis of their novels, you may combine your scoring of individual novels to characterize each novelist.

Third, while you are counting, it will normally be important to record the base from which the counting is done. It would tell us little that the word love appeared 87 times in a novel if we did not know about how many words there were in the entire novel.

Not all content analysis result in counting. Sometimes a qualitative assessment of the material is most appropriate.

Bruce Berg discusses "negative case testing" as a technique for qualitative hypothesis testing. Firsts in the grounded theory tradition, you begin with an examination of the data, which may yield a general hypothesis. Let’s say you are examining the leadership of a new community association by reviewing the minutes of meetings to see who made motion that were subsequently passed. Your initial examination of the data suggests that the wealthier members are the most likely to assume this leadership role.

The second stage is to search your data to find all the cases that would contradict your initial hypothesis. In this instance, you would look for poorer members who made successful motions and wealthy members who never did. Third, you must review each of the disconfirming cases and either give up hypothesis or see how it needs to be fine-tuned.

Let’s say that in your analysis of disconfirming cases, you notice that each of the unwealthy leaders has a graduate degree, while each of the wealthy nonleader has very little formal education. You may revise your hypothesis to consider both education and wealth as routes to leadership in the association.

This process is an example of what Barney Glaser and Anselm Strauss (1967) called analytic induction. It is inductive because it starts with observations, and it is analytical because it goes beyond description to find patterns and relationships between variable.

There are, of course, dangers in this form of analysis, as in all others. The chief risk is misclassifying observations.

Berg offers techniques for avoiding these errors:

1. If there are sufficient cases, select some at random from each category in order to avoid merely picking those that best support the hypothesis.
2. Give at least three examples in support of every assertion you make about data.
3. Have your analytical interpretations carefully reviewed by others uninvolved in the research project to see whether they agree.
4. Report whatever inconsistencies you do discover – any cases that simply do not fir your hypothesis. Realize that few social patterns are 100 percent consistent, so you may have discovered something important even if it does not apply to absolutely all of social life. However, you should be honest with your readers in that regard.

Several studies have indicated that women are stereotyped on television. R. Stephen Craig (1992) took this line of inquiry one step further to examine the portrayal of both men and women during different periods of television programming. Craig selected 2209 network commercials during several periods Between January 6 and 14, 1990.

"The weekday day part (Monday-Friday, 2-4 p.m.) consisted exclusively of soap operas and was chosen for its high percentage of women viewers. The weekend day part (Saturday and Sunday afternoons during sports telecasts) was selected for its high percentage of men viewers. Evening "prime time" (Monday – Friday, 9-11 p.m.) was chosen as a basis for comparison with past studies and the other day parts. "

Each of the commercials was coded in several ways. "Characters" were coded as:

All male adults

All female adults

All adults, mixed gander

Male adults with children or teens (no women)

Female adults with children or teens (no men)

Mixture of ages and gender

In addition, Craig’s coders noted which character was on the screen longest during commercial, as well as the roles played by the characters (such as spouse, celebrity, parent), the type of product advertised (such as body product, alcohol), the setting (such as kitchen, school, business), and the voice-narrator.

The table indicates the differences in the times when men and women appeared in commercials. Women appeared most during the daytime (with its soap operas), men predominated during the weekend commercials (with its sport programming), and men and women were equally represented during evening prime time.

Craig found other differences in the ways man and women were portrayed.

"Further analysis indicated that male primarily characters were proportionately more likely than females to be portrayed as celebrities and professionals in every day part, while women were proportionately more likely to be portrayed as interviewer/demonstrators, parent/spouse, or sex object/models in every day part…Women were proportionately more likely to appear as sex object/models during the weekend than during the day."

The research also showed that different products were advertised during different periods. As you might image, almost all the daytime commercials dealt with body, food, or home products. Instead, weekend commercials stressed automotive products, business products or services, or alcohol. As you might suspect women were most likely to be portrayed in home settings, men most likely to be shown away from home.

Other finding dealt with the different roles played by men and women.

"The women who appeared in weekend ads were almost never portrayed without men and seldom as the commercial’s primary character. They were generally seen in roles subservient to men (hotel receptionist, secretary, or stewardess), or as sex objects or models in which their only function seemed to be lend an aspect of eroticism to the ad."

Although some of the Craig’s findings may seem unsurprising, remember that "common knowledge" does not always correspond the reality. It is always worthwhile to check out widely held assumptions.

Strengths:

1. Economy in terms of both time and money
2. Safety
3. It permits the study of processes occurring over a long time.
4. It can have no effect on the subject being studied.

Disadvantages:

1. It is limited to the examination of recorded communications