There were sixty-three of them, written over nineteen years, addressed to her sister and never posted. The family found them after the funeral, in a shoebox at the back of a wardrobe, in chronological order and unsealed. The immediate assumption was that they represented a rupture: some grievance too large to send, held for two decades. The sisters had in fact spoken by telephone almost weekly for the entire period, and had holidayed together four times. Read in sequence, the letters do not describe a quarrel. They describe, at some length, the ordinary difficulty of saying anything true to someone you speak to constantly. Each begins with an observation the writer had evidently not managed to make out loud, and each ends, in most cases mid-page, without resolution. Several stop after a single paragraph. One consists of two sentences and a crossing-out. The temptation is to read them as confessions, and a few are. Most are not. They are drafts of a conversation that the telephone made both possible and impossible: possible, because the sisters were in constant contact; impossible, because constant contact establishes a register, and a register is hard to break. The letters are the residue of that difficulty. They exist because a weekly call has a shape, and some things do not fit into it. Whether the sister ought to have received them is the question the family has argued about since. The case for sending is obvious enough. The case against is that the letters were not, on the evidence, attempts to communicate. Nothing in them suggests an intended reader. They are addressed to a sister but written to an absence, and posting them would have imposed on a real person a document composed for an imagined one. There is a further consideration that the family has been slower to reach. The relationship, by every external measure, was a good one. The sisters were close, saw each other often, and by the account of everyone who knew them enjoyed each other's company without reservation. If the unsent letters are read as evidence of what was missing, they must also be read against nineteen years in which nothing appeared to be missing at all. It is possible that the letters were not a symptom of a failure but a condition of the success. The weekly call remained easy because the difficult material had somewhere else to go. This is not a comfortable idea; it suggests that a good relationship may depend on a certain amount of unsaid material being written down and put in a wardrobe. The surviving sister has read four of them and stopped. She keeps the box. Asked why she does not finish, she said that she recognised the voice in the letters but not the person it was speaking to, and that she was not sure she wanted to meet her.