Memory, the common (0) PRESUMPTION runs, is a kind of mental archive from which we retrieve events much as a librarian retrieves a book. The metaphor is appealing, not least because it the uncomfortable possibility that our recollections are anything less than reliable. The experimental literature of the past forty years has, however, away much of the metaphor's plausibility. Memories, it turns out, are not stored whole and then summoned intact; they are reconstructed, piece by piece, each time they are called upon — and the reconstruction is by everything from current mood to the specific wording of whoever is asking. Leading questions, as any competent defence lawyer will , can reshape a witness's recollection so that the witness may remain unaware that any reshaping has occurred. This is unsettling enough in the courtroom. It is perhaps more disquieting still in the ordinary of domestic life, where disputes about the past are generally settled not by evidence but by whoever sounds most confident. That confidence, the research suggests, is a particularly guide to accuracy. Indeed, it often in inverse proportion to it: the more certain a recollection feels, the more times it has likely been reconstructed, and therefore the more opportunities distortion has had to settle into apparent fact.