This talk will work with a corpus of candidate threats issued by parents in family mealtimes.  We identify a series of features of the threats, focusing on what appears to distinguish a threat from a directive, a warning, a complaint, an admonishment and other similar conversational phenomena. We will show how the different features of threats can be modified and exploited as the threats are intensified and undermined.  We use this preliminary analysis to pose a series of questions: In what sense can we talk about threats as dispreferred first pair parts? How far can one identify preferred or dispreferred next actions to a threat, e.g. compliance and defiance?  Are there alternative conversational upshots that can disrupt the effectiveness of the threat?  How are threats related to broader issues of asymmetry and socialization? Does it makes sense to think of a threat as a speech act?