This section describes the procedures used to code and analyze information collected from interviews. It should be noted that analysis was not a separate phase of the study to be implemented following completion of interviews. As has been done in previous studies (Finch et al., 1991; Lofland, 1971), the researchers chose to interview and analyze concurrently. Miles and Huberman (1984) also advocate analysis during data collection, stating that it allows interviewers to improve their craft. They maintain that such activity enables the researchers to "cycle back and forth between thinking about the existing data and generating strategies for collecting new--often better quality data" (p. 49). Through the use of this concurrent scheme, researchers were able to grow through experience and formulate a meaningful thematic structure for the final text from the collection of critical events.
As each interview was conducted, the interviewer completed a "write-up" for each incident. The "write-up" was prepared in the language of the interviewee, thus reading in the first person. It read like a story telling what actually happened. The purpose of the "write-up" was to organize and present the interview transcript and note-taking information in a more easily understandable sequence and format. Even though each interview was recorded and transcripts of the interview tapes were prepared, the "write-up" provided researchers with meaningful information that had been carefully organized to better facilitate coding and analysis (Mentkowski, O'Brien, McEachern, & Fowler, 1982). Information contained in the "write-up" was organized into sections which paralleled the Interview Protocol described in the earlier section on "Developing the Instrumentation."
A sample of an interview "write-up" is provided in Appendix A. The interview "write-ups" were checked against the original tapes of the interviews by another member of the research team in order to assure their accuracy and completeness.
After a few interviews had been completed, the research team met to discuss the coding scheme. Based on the initial Interview Protocol, actual experience from the interviews, the interview "write-ups," and using the conceptual structure of the LAI (Moss et al., 1994b), a coding sheet was prepared that integrated the potential attributes developed by the on-the-job incidents with the Interview Protocol. This coding scheme was to become the formal guide for interview analysis. As Boyatzis (1982) has noted, "the coding system attempts to explain how an interview should be assessed to determine the presence or absence, or degree of presence, of a particular characteristic" (p. 51). A coding system must, therefore, be clearly delineated so that various researchers agree as to how information will be coded.
While the initial Interview Protocol and the conceptual structure of the LAI were the starting points for the code list, other items were added so that all aspects of the interviews could be placed into a logical category. The decision to add new categories as needed aligns with the perception of Lofland and Lofland (1984) and Miles and Huberman (1984), who advocate modifying coding schemes during data collection and analysis instead of forcing data into predetermined coding schemes.
In addition to basic demographic coding for administrator code, interview code, gender, and type of employer, the following broad categories were coded for each interview:
The complete Interview Coding System sheet is provided in Appendix B. This sheet contains the finer detail of the outline under each of the elements listed above. The researchers agreed that a response should be coded in each of the eight categories, and that more than one response may be appropriate for some of the categories. However, a particular code was used only once per interview "write-up." In other words, while the interviewee may have provided several descriptions within an incident that would support the use of a particular code, the code was applied to that interview only once.
The researcher who conducted an interview did the first coding of that interview from the prepared "write-up." After the first coding of an interview, each interview was then coded a second time from the original "write-up" by a different member of the research team. This second person had also coded his or her own interviews prior to completing the seconding of a different set of interviews. Thus, each interview was coded twice by persons familiar with the interviewing, "write-up," and coding process. Both persons who had coded a set of interviews then met to discuss each interview. Where codes differed, the rationale for the coding was discussed until consensus was reached.
The coded interviews were entered into The Ethnograph V4.0 (Seidel, Friese, & Leonard, 1995) software package in order to obtain summary data about the frequency of each type of code and listings of the passages for each code type. Crosstabulations were carried out between major code categories to examine relationships between types of incidents and the leadership qualities which were developed. These summaries and crosstabulations are presented in the "Results" section.