Building A Classification Theory

1. Theory 2
2. Classificatory Theories
3. Explanatory Theories
4. Building A Classification Theory
5. Building An Explanatory Theory
6. Step 2: Identify causal factors
7. Step 3: Trace interactions among variables
8. Step 4: Propose a scheme for evaluating causal variables
9. Adapting And Revising Theories

Apparently the most common approach to constructing a classification system consists of gathering a quantity of data (items, cases, examples, responses) within a defined domain, then seeking to create categories (a) into which all of the items can be fitted (no items are left over) and (b) that will convey a useful meaning when the categories are compared.

To illustrate one way this can be done, we draw on our own experience with a project conducted to reveal the consequences people would recommend for the misconduct of individuals involved in six cases of wrongdoing. We use this example because we have an insider's view of the approach used.

The six cases varied from wartime crimes to adolescents' lawbreaking and to automobile accidents. Here is the target question the research was expected to answer:

In such moral incidents as those depicted in the six cases, what diverse rationales or types of moral reasoning will people adduce to support the consequences they would suggest for the wrongdoers in the incidents? ( Thomas & Diver-Stamnes, 1993, p. 1)

A more specific aim of the study was to compare the responses to the cases by six different groups of students, ones from: (a) a California innercity high school located in a socioeconomically depressed neighborhood, (b) a Catholic high school in a medium-sized California city, (c) high schools and universities in Saudi Arabia, (d) a junior college in Honolulu, (e) a high school in a small Hawaiian community, and (f) a teacher-education program in Northern California. A classification system would be needed to categorize respondents' answers so that the six groups could be compared. Because no established system was available to perform this task, we generated a system out of questionnaire replies from the 542 questionnaire respondents. The process was as follows:

We began with no preconceived organizational categories into which we would fit people's rationales for their suggested consequences. In other words, we chose to be led by the data, extracting categories inductively by listing the respondents' answers from the questionnaires.

As a first step, answers that were either identical or seemed clearly an alternate phrasing of the same idea were combined as one item in the list. For example, such responses as "Make the person suffer for the offense" and "The person should feel what it's like to be hurt" formed a single item. Even with such combining of answers, the list was very long, extending to more than 200 items. In effect, although respondents were reacting to only six moral incidents, they provided a great diversity of moral-value rationales.

As a next step, the items that appeared similar in intent were conceived to form a cluster, and a word or phrase was created to reflect the apparent essence of that cluster. To illustrate, the two responses "She was too young to know better" and "In the society he comes from, what he did was not considered improper" were both placed in a category entitled Awareness of Right and Wrong".

Finally, the labeled clusters were cast into five major groupings that appeared to represent broader generalizations than did the clusters. For instance, clusters labeled Prevent Future Offenses by the Actor, Deter Others from Misconduct, and Wreak Revenge were collected under the title The purpose of the consequences. This process resulted in five overarching categories that represented the general framework for the taxonomy, (1) moral values, (2) purpose, (3) conception of causality, (4) consequence feasibility, and (5) agent qualifications.