A study explored how Retrospective Miscue Analysis (RMA) is successful, using discourse analysis of the questions and discussion strategies used during RMA. Seven readers with successful RMA sessions who represented several age/grade groups, ethnic groups, and linguistic backgrounds were chosen, and their discourse was analyzed over a period of 8 months. Examining what aspects of the RMA session support the readers' development of a positive view of himself or herself as a reader, over time, 4 subcategories emerged: procedural, language used, interactional, and ownership/control. Examining also what language the reader uses to talk about his or her reading process and strategies, over time, the study found 2 subcategories: reading process and metacognition. Findings show that a common thread through all of the readers' RMA sessions is their movement toward central roles in the dialogs, toward equanimity and parity with the teacher/researcher's contributions. This increase in discussant role goes hand-in-hand with an increase in the ability to talk about reading. As readers become more fluent in the language of reading, their confidence in their ability to learn about the reading process grows, and they become more in control of the RMA sessions. It is precisely this control over the reading process which influences reading proficiency. Teacher/researchers also learn how to discuss the reading process with each individual reader. RMA sessions tend to progress from a primarily teacher/researcher-driven dialog in the beginning sessions to a conversation between equals in the middle and latter sessions, during which it is frequently the reader who drives the session. (Contains 14 references, and 2 tables, and 13 figures of data.) (SR)