The review was commissioned to establish who had approved the change. It concluded, after nine months, that nobody had. This was not a case of concealment. Every meeting had been minuted, every document retained, every participant willing to be interviewed. The difficulty was subtler and, for that reason, harder to write up. The change had emerged from a sequence of individually reasonable steps, none of which constituted a decision, and the sum of which nonetheless produced one. The reviewer traced it back. A working group had been asked to consider options. It produced a paper listing four, with a preference for the second. A subcommittee, reading the paper, took the preference as settled and asked how the second option might be implemented. The implementation note, drafted for that narrower question, described the option in the present tense, as departments tend to do. By the time the note reached the board, the language had hardened: what had been a preference among four was now *the approach*, and the board, seeing no dissent recorded anywhere, approved the timetable rather than the substance. Nobody had been disingenuous. Each step was defensible in isolation, and each participant could point to a document showing that the decision had been taken elsewhere. The reviewer, whose brief had been to identify accountability, found herself instead describing a mechanism. Her report is unusually candid about the implications. Organisations are designed on the assumption that decisions are events: they happen at a moment, in a room, and can be attributed. Much of governance follows from this assumption, including the entire apparatus of minutes, sign-off and audit. But some decisions are not events at all. They are the residue of a process, and the process has no author. She was careful not to present this as a scandal. The change itself was probably right; two years on, it appears to have worked. The problem is not that a bad decision escaped scrutiny. The problem is that a good one did, and the organisation therefore learned nothing it could apply again. Her recommendation was correspondingly modest, and was not adopted. She proposed that any paper listing options should be required to state, in a single line, whether a choice among them had been made and by whom, and that the absence of such a line should itself be recorded. It is a small procedural change and would have cost almost nothing. The board considered it at a meeting in November. The minute records that members welcomed the report, noted its findings, and asked officers to reflect further on the recommendation in the context of the wider governance review. There is no record of any subsequent consideration. Whether the recommendation was rejected, deferred, or simply lost is, in the circumstances, an appropriately difficult question to answer.