Preserve History Without Negotiating With It

A working file in which every deleted option survives in brackets, alternates and commentary, “[or: she turns away],” “[earlier version of this paragraph below],” each kept because deletion felt irreversible, no longer shows the current state of the Work. A file like that is hard to read, so it is hard to revise. The opposite failure is revising timidly because there is no safety net. Keeping what was without letting it govern what is can sound like two instructions in conflict.

A checkpoint resolves both. Before a bold revision, make an identifiable copy of the manuscript, with a date and a one-line note on what you’re about to try, and after that revise the live file without keeping the dead alternatives inline. A dated folder, a commit in a version-control system, or a saved version in whatever writing application you use can all do the job; the discipline of separating the record from the draft matters more than the technology. Part of the reason is probably psychological. A change feels safer when it can be reversed, and you can revise more boldly once you’ve stopped calculating whether to undo it.

Fitzgerald’s galleys for the novel then titled Trimalchio show what goes missing without one. Princeton’s archive holds the autograph manuscript and the corrected galleys; according to its curators, the typescripts that stood between them do not survive, and what happened between those two states cannot be reconstructed from the archive. That is the gap a dated checkpoint closes. The scholar Matthew Bruccoli, quoted in Princeton’s account, argued that Fitzgerald regarded galleys as a special kind of typescript or trial edition in which to rewrite whole scenes as necessary.

A cuts file, the separate document where removed passages are kept (Practice 84), is a legitimate resource, and earlier versions may contain a sentence worth restoring, or evidence of when a decision was made. More versions are not always better, though. A writer who keeps forty numbered copies and compares them compulsively may be negotiating with the past instead of deciding.

Deleting a passage from the current manuscript does not delete it from the history. If a file will go to a reader, an editor, or a collaborator, earlier text can travel with it, in tracked changes, comments, saved versions, or file metadata, depending on the software. Before sending, export a clean copy and look at what is actually in it. This matters most for anything you’ve deliberately removed for the sake of privacy, or on behalf of a real person.

Some collaborative or translated work legitimately carries live alternatives. An editor’s tracked changes are meant to sit inline until the author resolves each one, and a translator may carry two renderings of a line until a decision is possible. Those alternatives are pending rather than dead, and nothing pending should remain when the manuscript is declared current. To tell a live alternative from a dead one, ask whether anyone is going to choose between them.

Practice. Create a named checkpoint of the manuscript, with a date and a note. Then spend thirty minutes revising one section decisively, deleting dead alternatives and leaving no brackets and no commentary in the text. Wait a day, compare the section against the checkpoint, and confirm that what was removed is what you meant to remove.