Study Design, Methodology, and Profile 33over the “reality” of the marriage and divorce at issue.^6 I build on these method-
ologies, adding quantified attention to race and gender, so that we can better track
the kind of patterning Conley and O’Barr indicated might be emerging from their
study.^7 In this I draw on the work of Matoesian, Merry, and Frohmann, who use
discourse analysis to trace class- and gender-based differences in orientations to-
ward the law.^8
The resulting study, then, is a broader picture than could be obtained by study-
ing one or two classrooms and permits us to begin to map possible patterning of
law school discourse. The classes are not to be viewed as a random sample, of course,
but as a set of in-depth case studies that can be compared with one another and
with other existing studies of law school classrooms.
Access, Site Selection, and Project Process
The eight schools selected for the study were chosen in an effort to maximize di-
versity of school status and professor profiles. In many cases, a combination of the
issues involved in gaining access to classrooms and availability of proficient cod-
ers also shaped the choices. Access to classrooms was sometimes quite difficult to
obtain; in several instances, professors declined because they understandably felt
self-conscious about being observed, although the deans of the law schools involved
had granted permission. In one case, indeed, we were asked to provide a precise
description of the discourse features we’d be analyzing; this we declined to do be-
cause of possible reactivity effects (i.e., that professors who were aware we were
observing certain dimensions of classroom interactions might become self-
conscious and alter their approach to those aspects of the discourse). Another lim-
iting factor was the availability of suitable coders; only coders who had graduate
training in linguistics, anthropology, or sociology were used for the study; in one
case we had to drop a particular location because we couldn’t obtain coders who
met these requirements. The combination of selection for diversity, professor ac-
cess, and coder availability shaped the choice of research sites.
I personally observed and coded a full semester of classes in one of the schools
of the study; the remaining schools were handled by other observers. Each classroom
was covered primarily by one coder (although coders also arranged for coverage by
a backup person who could tape and code on days when they were ill or otherwise
unavailable). This ensured a high level of within-classroom consistency in coding
for each school; in addition, it raised overall coding accuracy. Perhaps not surpris-
ingly, growing familiarity with the students’ identities and professorial style meant
that coders with more experience in a particular classroom could follow rapid-fire
exchanges without missing a beat. Thus, in several cases, when I flew in to perform
intercoder reliability checks (I independently coded classes alongside the coders and
then cross-checked our results), I found that despite having pioneered the coding
method myself, I missed a turn or two that the regular coder for that classroom caught.
(Fortunately, subsequent cross-checks of all coding using tapes and transcripts en-
sured that we would in any case eventually catch any missed turns.)^9
I also conducted interviews with small focus groups of students from the class-
room that I coded and interviewed six of the eight professors who participated in