Developing new Inquirer Categories — Basic Strategies

Most of the issues to consider in developing a new category are examined in the first two chapters of our MIT Press book, The General Inquirer. We still recommend these chapters and support their recommendations.

A key issue in developing categories is whether the category is to consist of words and word senses that describes a concept, or whether it consists of the language used to express that concept. For example, a category may consist of different synonyms, broadly considered, for "achievement". This may be quite different from a category of the words with which a person expresses an achieving orientation. Some categories may in fact be a mixture of both, but this may not be as productive as having clearly separated categories.

There also may be several different ways that a concept is expressed. For example, people may "flame" in their email in several different ways and it may be helpful to have separate categories that distinguish between these different ways that people flame. Similarly, there might be different ways of expressing protest that can be represented by different categories. Content analysis can be helpful to sort out and identify these differences.

To develop categories, it is usually helpful to start with documents that are from known sources to see if they sort out into different language groups. For example, the documents may all come from people who have been rated as being high in achievement. The question then is how this achieving orientation shows up in the language they use. And does it always appear as one type of language, or are there several achievement orientations, each with its own language?

Another way to develop such categories is to compare documents that address different objectives. For example, protest documents can be addressed to seek change or redress or termination of a behavior and the language in each case may be different.

If there are clearly grouped documents, the "word" option in the Java Inquirer program may be used to obtain words counts of all words and words senses in the Inquirer dictionary. The investigator can then enlist discriminate-function procedures to separate the documents on these basis of these word counts into clusters. Some of the clusters may be meaningful for identifying further distinctions about the concept. But because this is problematic (especially if there are a limited number of documents or there are other differences between documents not related to the concept being investigated) some category developers may prefer to look directly at the count patterns and use this information in developing their categories.

Being able to process a large amount of information opens up new opportunities to build useful categories. One may start with a priori notions of what a category should contain. But later expansion on the basis of patterns in how people actually talk can only help make for more realistic and valid content-analysis research.

Return to Home Page