This conversation analytic study shows that well-prefaced, self-initiated repairs in talk-in-interaction manage the plausibility of claims in talk, more specifically the relevant precision of a turn-so-far by reconsidering it yet without withdrawing it.