Methodological Procedure: Huckin

1. Pose a Research Question. Content analysis proceeds best if the researcher has a good research question to start with, one that: (a) addresses a topic or issue of likely interest to fellow members of the researcher’s discourse community, and (b) constitutes a novel claim about this topic or issue…A research question becomes a hypothesis if it is sufficiently explicit to be tested and, if appropriate, refuted by empirical data. A deductive study requires a hypothesis, an Inductive study does not.
2. Define the Appropriate Construct(s). A good research question will focus on one or more general concepts such as “writing quality” or “bias.” Such a concept is the construct of interest for that study (MacNealy, 1999). If the research question is worded in such a way that the construct of interest is not obvious, the researcher must make a special effort to clarify it…
3. Select an Appropriate Text or Body of Texts as the Study Corpus. An appropriate study corpus is one that will provide a good test of the research question, which means (a) that all the texts included in the corpus correspond in some clear fashion to the research question, and (b) that these corpus texts are representative of some identifiable, larger body of texts…A project aimed at identifying only a small set of surface variables, for example, requires less time and effort per text than a project concerned with deeper, more complex variables; thus, a project of the former type should use a larger corpus than a project of the latter type.
4. Determine Appropriate Units of Analysis (Text Features), UsingMultiple Raters if Possible. The units of analysis, or text features, to be quantified for content analysis must be those that emerge logically from the research question. They must have a direct bearing on the question, in the sense that the incidence of a feature will constitute either direct support or direct nonsupport for it. The categorization of these units of analysis should be readily identifiable and non-overlapping, so as to be codable with some degree of reliability…Working collaboratively with one or more other investigators can help resolve uncertainties and produce a sample list of reference terms…Once the units of analysis have been determined, two or more investigators should independently sort at least 10% of the data into categories, and then the respective sortings should be compared for interrater reliability.
5. Gather Data. If the units of analysis have been narrowly defined in Step 4, data gathering should be a relatively straightforward matter of identifying and counting.
6. Interpret the Findings. The final step is to analyze the data against the research question, that is, to interpret the findings…In an inductive study, the research question is only tentative and may be in need of revision if it does not conform to the data. Indeed, the goal of such a study should be to construct and refine a research question that might stand up to a more deductive, follow-up study.