This paper argues that explanations do various things, and that the analyst would do well to look to the explainers themselves, and their audience, to identify just what those things are. As an illustration of the general principle, the paper describes explanations that need to be understood as arising from speakers' construction of each other as being accountable. ‘Explanation slots' can be set up for another speaker by various ways of signalling something problematic in what they do or say (or fail to do or say). Such slots can also be set up for oneself, in which case they serve as markers of one's own behaviour as non-normative and requiring explanation. To show how this setting up and filling of explanation slots is deployed in the pursuit of local projects at short or long range, I apply the analysis to a sample interaction between a candidate and two interviewers. The promise of the conversation-analytic approach taken here is that it liberates the researcher from reliance on a priori categorizations of the content of explanations (for example, as causal attributions or self-serving accounts), and looks instead to the practices of interactants themselves as the grounds for determining their meaning. Moreover, it integrates ‘explanations’ into the fabric of what is happening in the interaction, rather than treating them as free-standing products of cognitive processing or psychodynamic motivation.