NEITHER RIGOROUS NOR OBJECTIVE? 71craft, rehearsing a repertoire of patterns of action and interaction and response for months before
going on stage, so, too, interpretive researchers learn the action repertoires of their research craft:
how to select “good” research sites—places where they will be more likely to observe what it is
that they want to see; how to identify “good” documentary locations or “good” people to chat
with; how to “topic talk” with them; and so on. One develops a repertoire of reading, conversa-
tional, and/or participatory “moves”—ways of framing a question, of following a lead, of re-
sponding to what has just been said, of joining a group engaged in an ongoing conversation (see,
e.g., Walsh 2004), of getting oneself invited to meetings, and so on. As with theater, research
“improvisation” consists of drawing on that repertoire as the specific situation requires, in the
context of one’s research role and in keeping with one’s personal values: responding appropri-
ately to what someone has just said (“Yes, I had to find a back-alley abortion quack”), to an
invitation just delivered (“Hey, let’s go shoot some dope”), to a heretofore obscure reference
(“File note to myself: Dig up Kerry’s Congressional testimony”).^7
Such replies and acts cannot be scripted in advance, as one cannot anticipate what others will
say or do; and so, as Katz (2004, 7) notes, this lack of a preestablished research protocol with
detailed time- or place-based steps is “not a sign of the [researcher’s] regrettable incapacity to
plan research.” Rather, it means that interpretive research cannot be “strict,” without deviation
from such a plan; but being adaptive to human response does not mean that the research cannot
be, or is not, systematic. This inability to script the research ahead of time relates to another charac-
teristic of interpretive research: the fact that one typically does not start out with developed hypoth-
eses, which one then tests with “field” data, but rather allows one’s hypotheses or explanations to
emerge from extended immersion in the data themselves. Interpretive research rarely proceeds from
a formalized hypothesis because the researcher does not know ahead of time what meaning(s) will
be found, expecting them to be generated through (participant-) observing and/or conversational
interviewing and/or the close reading of documents. The general “hypotheses”—“hunches” would
be a term more characteristic of interpretive procedures—carried by the researcher into the data-
generating phase or site are more likely to concern ways in which meaning(s) is (are) communicated
than to be specific ideas about things to be explained in the setting under study. Generalization, too,
is more likely to concern communicative, meaning-making processes than substantive rules or prin-
ciples. Interpretive researchers are quite serious about letting the data “speak for themselves,” resist-
ing the impetus to rush to premature judgment and analytic closure, and this limits their ability to
prespecify operative variables and their measurement. This is the idea that Glaser and Strauss (1967)
sought to capture in “grounded theory” see also Locke 2001; Strauss and Corbin 1990). Its very
unscriptedness makes writing proposals for interpretive research problematic for granting agencies
and committees that see science as always being hypothesis driven.
Modes of analyzing interpretive data are less “improvisational,” but here, another characteristic
makes it appear on the surface as if they lack systematicity: It can be difficult to make explicit how
one goes about making sense of one’s data. Formal semiotic analysis is perhaps the most stepwise,
procedurally, of all interpretive analytic methods: The steps entailed in setting up semiotic squares,
clusters, or chains can be articulated clearly (see, e.g., Feldman 1995; Gottdiener and Lagopoulos
1986). But even here, discovering the connective tissue of meaning often comes as much from a
flash of insight as it does from following any elaborated system of rules. As Feldman (1995, 30)
emphasizes, it would be more accurate to see the analytic device (e.g., the semiotic square) as a tool
for stimulating thought than as itself producing that thought (unlike, for instance, a regression analy-
sis in which the computer generates the findings in the form of the coefficients and R squared).
The most exacting descriptions of forms of interpretive analysis describe a kind of indwelling
with one’s data: whether using index cards held in the hand or large sheets of paper tacked to the