This corpus-based investigation explores the features and discourse distribution of <i>it</i>-clefts. It is argued that their core meaning is the positive identification of a discourse element, usually a subject/agent NP, mostly by means of the declarative pattern <i>it is/it was . . . that </i>in spontaneous persuasive speech. The construction is found to implement three main strategies: (a) corrective when reformulating old topics, generally displaying the pattern new + given; (b) transitional when (re)introducing new or deactivated topics or spatio-temporal settings, usually with the pattern new + new; and (c) topical when continuing with a previous discourse topic, normally exhibiting the values given + new. Lastly, the central function of <i>it</i>-clefts is shown to be that specifying a relation of exhaustive topicality.