INTERPRETIVE CONTENT ANALYSIS 335that the researcher start with “the conventional definition of stories,” and he defines stories as
having “beginnings, middles, and ends, as in scenarios” (Roe 1994, 3). As noted above, one of the
main tasks that agency analysts pursued in producing the wilderness EISs was to develop future
scenarios about the public lands and resource use and protection. Analyzing the documents to
identify their story lines reveals how the EISs contributed to the framing of wilderness policy
inside the agency.
Roe (1994, 3) indicates further that “if the stories are in the form of arguments, they have
premises and conclusions.” Majone (1989, 23) asserts that “practicing policy analysts often en-
gage in argumentative discourse [as they] debate values, question objectives, agree or disagree
about assumptions, and advocate or justify courses of action on the basis of less-than-conclusive
evidence.” In his estimation, “argument is the link that connects data and information with the
conclusions of an analytic study” (Majone 1989, 10). In examining the wilderness EISs, I consid-
ered the form of arguments they contain in terms of premises and conclusions, how evidence
(e.g., data and information) is linked to conclusions in them, and how the arguments they contain
draw on normative positions about wilderness designation of public lands. Analysis of arguments
in these terms provides a systematic means to show how the EISs contribute to framing the policy
issue of wilderness designation.
In summary, I examined how agency analysts interpreted and framed the issue of wilderness
in the EISs through stories and arguments that link technical information to policy recommenda-
tions. Because the documents exist in both draft and final form, I also considered whether there
existed evidence of changes in the framing of the wilderness issue between earlier and later ver-
sions of these documents. Changes in the framing of the wilderness issue provide evidence about
policy change for public lands management more generally.
Before describing components of the analysis, I comment on how I arrived at the point of
conducting an interpretive analysis via a project that involved statistical analysis of data and
recommendations in the EISs. The interpretive analysis was, in part, a response to that project.
FROM STATISTICAL TO INTERPRETIVE ANALYSES
At the same time that I began my ethnographic study, I was hired as a research assistant for a
project at the University of Michigan to help analyze a set of wilderness EISs produced by the
BLM. The objective of the analysis was to determine whether there existed statistically significant
relationships between technical information in the documents and agency recommendations for
designations of wilderness areas. This project led to a disjuncture that helped to motivate and
inform my interpretive analysis of the EISs.
In conducting the statistical analysis of the wilderness EISs, I coordinated the efforts of a
group of students to code information and policy recommendations from the documents. I worked
with the principal investigator on the project to conduct statistical analyses. The academic litera-
ture on EISs included questions about whether and how technical information in them is linked to
decision making. We tested for such relationships using regression analysis. We reported that this
analysis indicated that very little of the technical information was related, with any degree of
statistical significance, to BLM policy recommendations. We concluded that this lent support to
the idea that the BLM had produced these EIS documents primarily to fulfill a legal mandate
rather than to inform substantive policy decisions (Ginger and Mohai 1993).
This conclusion generated a small crisis for me. On the one hand, I was gaining valuable
research experience and helping to produce outputs valued in academia. On the other hand, our
conclusion did not reflect the complexity that I was experiencing as a participant-observer in the