The thesis uses (semi-)open-ended interviews with drug users and non-drug users to document the lived practices of qualitative interviewing and talk about drugs. It draws on Sacks and more contemporary work on conversation analysis and membership categorization to outline in detail 'how qualitative interviews come off in and through talk-in-interaction. It also contextualises the interviewees' topic talk, their talk-about- drugs. An (intimately) related concern, is to document some ideals about qualitative interviewing and talk about drugs as ideals-in-and-as-lived-practice. Underlying all types of research interviews is the tension between an extra-local need to collect data on a topic and a here-and-now interactional event in which this data is collected in and through talk-in-interaction (Antaki and Rapley 1996, Mazeland and ten Have 1996/1998, Suchman and Jordan 1990). The thesis describes how interviewers and interviewees manage this tension, documenting (some of) the methods - the practical solutions -they routinely use "to get the job done". Special reference is given to how interviewers locally produce themselves as 'the sort of qualitative interviewers they are supposed to be'. The research outlines how qualitative interviews are locally produced, with reference to the structural, sequential and topical organization. It focuses on (some of) the methods interviewers use explicitly to inform the interviewee that their questions are to be heard and understood as neutralistic and/or facilitatory. These methods include producing questions without preferred responses, question prefaces, specific lexical choices and tag-components. It also outlines (some of) the methods interviewees use to produce themselves as 'morally adequate' types-of-people in relation to topics in the talk. This thesis seeks to unsettle some of the current research practices and theories in the academic 'drug' and 'qualitative interviewing' worlds. Chiefly, it shows how both speakers, the interviewee and interviewer, are essential to, in Silverman's (1973) term, 'bringing off the research instrument'.