This article considers which people talk about important matters, what people talk about when they discuss "important matters," and the implications of conversation topic for the interpretation of results arising from the General Social Survey (GSS) network instrument based on the "important matters" name generator. We show that half the people who report not talking about anything have nothing to talk about, whereas the others have no one to talk to. Secondly, we show that people tend to talk about things that many would regard as unimportant, for example, cloning of headless frogs, eating less red meat, and so on. Given this, the connection between characteristics of discussion networks and achievement of instrumental ends — for example, getting a job or enhancing social support — is tenuous. Finally, we show that there is substantial topic-alter dependency. This dependency suggests that many substantive findings reported about, for example, gender differences in network composition might be an artifact of the data-collection instrument. Micro-level topic-alter dependencies reflect macro-level associations between attributes, topics, and roles. Consequently, cross-cultural comparison of GSS network questions is problematic. Solutions for escaping these methodological dilemmas are proposed.