The Vault, the Ledger and the Vanished Film

AA film is a chemical event before it is a work of art. Every reel in a vault is a strip of plastic slowly returning to its constituents, and the archivist's first duty is not interpretation but delay. Cellulose nitrate, the standard support for professional film until 1951, decomposes into an acid that attacks the image it carries; the acetate that replaced it fails more quietly, warping and giving off the smell of vinegar that has given the process its name. Neither failure can be reversed. A survey published in 2016 by the Federation of European Film Vaults estimated that of the 12,300 features produced in Europe before 1930, fewer than 2,100 survive in any form, and that barely a third of those survive complete. Figures of this kind appear in every funding application, but they conceal the harder question, which is not how much has been lost but which of the surviving copies an archive should be willing to call the film.

BTwo answers compete for the same budget. Photochemical preservation copies the image onto fresh polyester, a stock its makers guarantee for five hundred years in a vault held at five degrees; the copy can be read by anyone with a lamp and a lens, and it asks for nothing further. Digital preservation scans the frames, most often at 4K, and yields files that are cheap to circulate and expensive to keep, because storage formats must be migrated roughly every seven years or become unreadable. Between 2009 and 2021 a consortium of 34 European archives raised the number of features scanned at 4K from about 400 to more than 5,600. What did not rise was the money for the vaults: spending on refrigerated storage across those same institutions stayed within three per cent of its 2009 level throughout. Marta Kowalczyk of Lodz, who has directed several of the consortium's projects, states the consequence flatly: 'we learned to copy faster than we learned to keep'.

CKowalczyk's own argument rests on a corpus of 214 restorations completed since 1995. Among restorations of films for which a laboratory day book survives, the ledger in which technicians recorded tinting baths, printing lights and running speeds, 71 per cent departed from the appearance of the best surviving print, usually by returning colours the print had lost; where no ledger survives the figure is 12 per cent. Kowalczyk reads that gap as a sign of restraint rather than of licence: restorers alter the visible copy when a document tells them what the film once looked like, and leave it untouched when nothing does. On this reading a restoration is an argument supported by paperwork, and an archive's obligation is to publish the ledger beside the film so that the argument can be tested. A projected image, she has written, is a hypothesis about the past, and a hypothesis is worthless unless the evidence for it is placed where others can attack it.

DDuncan Reeve, of Bristol, accepts the measurements and rejects the inference. In his account the day book does not drive the decision; it licenses one that has already been reached, since a restorer who wants a saturated image can find in almost any ledger a figure that supports him, while a restorer trained in a soberer decade will find the opposite. He then presses a harder objection. The prints that reach an archive are not a sample of what was released but of what travelled, and the copies that travelled furthest were cut, retitled and regraded for each market, so any correlation between documentation and intervention may record the habits of distributors rather than the intentions of directors. Reeve concedes two points. His comparative material comes from a single country, Britain, and the restorations in it are dated largely from festival programmes rather than from the laboratories' own dockets, a source he elsewhere calls the softest ground in the field.

EThe exchange has narrowed on two points. Kowalczyk accepts, in a concession much quoted against her, that the ledger correlation is confined to films made before synchronised sound arrived: among titles first released after 1930 the difference between the two groups is under four percentage points, and in the 1950s it runs the other way. She also grants that the software now used to manage grain can settle nothing between the two explanations, since every laboratory applies it to every kind of material. Reeve answers that the silent figures are the interesting ones precisely because the studios of that era kept dockets, and that those dockets should now be read against the censorship files recording what audiences were actually shown. Neither position has been abandoned. Since 2015 most major archives have refused to project original nitrate at all, which means the dispute will be settled, if it is settled, by surrogates that carry no evidence of their own making.

FOutside the conference hall the argument has a price. In 2018 a restoration was presented at a festival as the director's cut of a melodrama of 1927; a technical report the following year identified an optical soundtrack negative made in 1953, the year the film was reissued. The report fixes the date of the sound and says nothing about the date of the images, which the tinting evidence still places in the 1920s, a distinction that survived in the archive's catalogue but not in the reviews. Funding bodies now ask for a provenance report as routinely as they once asked for a director's name. Kowalczyk regards this as the clearest gain of the past decade; Reeve, who has written two such reports, warns that a method able to expose a later element is not thereby able to recover an original intention, and that it is the second claim on which festivals sell their tickets.