Focus group discussions are now widely used for gathering data, in soc
ial science as well as in commercial marketing and public opinion rese
arch. One appeal of focus groups is that in some ways they seem like e
veryday talk, but their effectiveness depends on a tension between the
moderator's constraints and participants' interaction. The moderator
introduces and defines topics, but participants can shift, close, and
interpret them. The moderator elicits disagreement in a way specific t
o focus groups, but participants manage their disagreement. Thus we se
e not simple control by the moderator, but a complex collaborative pro
ject operating under the shared assumption that the purpose of the dis
cussion is to display opinions to the moderator. These findings extend
the analysis of conversation in institutional settings and contribute
to a methodological critique of the reification of attitudes and opin
ions in some social science research.