When the Research Disproves the Film You Wanted to Make
When the Research Disproves the Film You Wanted to Make
The entry is four lines long. A date, a fault, a repair, a signature. Someone wrote it in a maintenance log because the work had been done and someone would eventually want to know what it cost. Nobody wrote it for you.
You came to the archive looking for texture — a date to hang a scene on, a name, a photograph — and instead you found the sentence your film rests on, contradicted. Not shaded. Not complicated. Undone. The clock your treatment describes as having run without interruption for a century stopped, and here is the paper that says so, in ink, in a book that has been in the same building the whole time.
What follows is a specific temptation, and it is worth naming before it does damage: the search for a way to keep the film. A footnote. A line of narration conceding "a brief pause for maintenance in 1934" before returning to the unbroken sweep of the hands. The engine has been lifted out of the thing and you are pricing a way to keep driving.
The shorter answer than the panic suggests: write down the exact claim that failed, work out what it was holding up, and then choose among questions you can actually research — or stop, and say so out loud. Research that breaks a premise has to change the film's argument and its material plan. A disclaimer cannot carry that load, because a disclaimer leaves the old sequence standing and asks the audience to hold a contradiction the film hasn't resolved.
The case I'll work through below is invented. It's a paper comparison built so I can carry one set of conditions through every route without attaching a real correction to a real town and a set of records I haven't inspected. A documented filmmaker whose premise collapsed under identifiable evidence would be better authority here, and it's worth noting why that's a hard thing to cite: the account of a decision and the verification of the facts underneath it are two different pieces of evidence. "I dropped my film after I found X" is testimony about a choice. It does not establish X. Hold that distinction; you'll need it in your own project too.
Establish what the new evidence actually contradicts
Three questions, in order. What does the record say. Who made it, and why. Does it reach the claim.
The claim, written plainly. Not as a mood, not as a theme — as a proposition that is either true or false. "The clock has run without interruption for a hundred years." Once it's in that form you can see how much is being asserted. The absolute is doing the work. A weaker version — "the clock has kept the town's time since 1898" — says something about function, not continuity, and the record may not touch it.
The record, read carefully. Maintenance logs exist to schedule work and account for money. That makes them excellent evidence of the events they record and poor evidence of the events they omit. An entry showing the clock found stopped on a Tuesday morning proves a stoppage. A quiet stretch of pages proves nothing at all, and you should resist reading it as a good sign. The asymmetry matters: your record can falsify the strong claim on the strength of a single line, and cannot confirm the surviving weaker one. You will spend the next month wanting the second thing and only ever getting the first.
Then check whether the entry falls inside the period the claim covers. One stoppage inside the hundred years kills the hundred years, regardless of whether the log begins in 1934 or 1898. That's the rare mercy in this situation: a narrowed record can still be decisive.
Terminology, which is where most false alarm lives. "Stopped," "lost eleven minutes and was regulated," "taken down for cleaning," "fitted with a new motor," and "run by a different movement entirely" are five different facts. Our invented log has an entry from 1961 for the movement being removed for cleaning, during which the clock did not run, and a filmmaker whose first instinct is to argue that a scheduled removal isn't a stoppage. The record does not care. But you should notice what you're doing in that moment, because it's the same move that will later show up in narration.
Provenance, scope, limits. Who kept the log and under what obligation. Whether it's the complete run or the portion that survived a flood. Whether a second record — a council minute, a bill of sale, a keeper's diary — explains an entry that looks damning. And whether the absolute claim was ever made by anyone authoritative, or whether it was your own compression. If the historical society's plaque says "the clock has served the town since 1898" and your one-pager upgraded that to "has never once stopped," the contradiction is entirely internal. That's the least dramatic version of this article's subject and the most common one, and the correction is easier, and it still has to be made.
What you must not do is announce disproof before doing this inspection, and what you must not do afterward is manufacture a dispute. If the record settles the question, "some say it ran continuously" is not a second side. It's a sentence you'd be writing to protect a premise.
Separate the broken premise from the remaining subject
Before you go looking for a replacement angle — before, because otherwise you will design the replacement around the footage you already have rather than around what the evidence supports — draw the map.
Scenes that exist only to serve the claim. The time-lapse that needs the hands to complete an uninterrupted sweep. The drone move over the town that means "a century of steadiness." Anything requiring the clock to be a symbol of continuity rather than an object in a tower.
Proposed interviews that were chosen because of the claim. The third-generation keeper brought in as the man who has never had to do anything. Ask what you'd want from him if the premise were gone, and whether you'd want him at all.
Character descriptions that encode it. "The custodian of a machine that has never needed him" is a lovely line that is now fiction.
The emotional promise. This is the big one, and it was already arguing with itself. The original film was going to say that continuity is a virtue, that a town's endurance is legible in a mechanism, that care and permanence are the same thing. The premise underneath it said the opposite: the clock had needed no care, which made the endurance seem effortless and therefore noble. The narration avowed one relation between care and permanence; the premise delivered the other. The record takes the avowed side. People have been doing unglamorous work for a century to produce an appearance of effortlessness, which means the thesis survives in a form you can now defend — and the feeling that carried it does not. That is a different and arguably better subject. It is not the same film.
What remains supported independently is unglamorous and real: the tower, the clock, the workshop, the trade, the ledger, the money spent. A compelling setting survives the collapse of an origin story. That comfort has a sharp edge, though — "the tower is still beautiful" is not a film, and the fact that you can still point a camera at something is not evidence that a film about it is warranted.
Compare reframing with stopping, not only with minor repair
There are four honest routes here, and three of them are harder than they look.
A narrower question. New question: what does it take to keep a public clock going for a hundred years? Usable material: the maintenance records read in full — what work they describe, not just when — plus workshop access. Unresolved: whether the records describe process at all, or only events and costs. A log that says "repaired, £4 6s" gives you a paper trail but not a scene. You may also find the workshop is a place with nobody in it, which is an access problem you can only discover by asking.
An inquiry into how the account developed. This is the route that feels most like a revelation and is most likely to be a research project wearing a film's clothes. New question: where did "never stopped" come from, and what was it doing for the people who repeated it? What it requires is reporting you have not done and cannot currently support: guidebooks, centenary programs, tourism copy, minutes, plaques, the earliest printed instance of the phrase. Right now, in our example, the record contradicts the claim and says nothing whatsoever about the claim's career. Those are separate facts.
Two hazards. First, the reporting may come back empty, in which case you have spent months and have neither the original film nor this one. Second, the film this route becomes can curdle into a smirk — a documentary about how a small town got its own history slightly wrong, which is cruel in a way that sneaks up on you. It can also curdle the other way, into invention. Absence of an explanation is not evidence of a cover-up. A phrase can drift from "it kept going through the war" to "it never stopped" across fifty years of retelling with no liar anywhere in the chain. If you cannot document who said what, you do not get to cast the parts.
A different evidenced relationship to the clock. The records that killed your premise are themselves material. What they document besides stoppages — parts, suppliers, marks, wages, names — is something you learn by reading them, not by deciding it in a treatment meeting. New question: whatever the record actually supports. Unresolved: you cannot pitch this until you know, and knowing takes weeks.
Stopping. Not a failure of nerve and not a moral position. A real option with real conditions: the premise was the film — it supplied the sequence, the tone, and the money. Every remaining route requires new reporting with uncertain returns. The access you had was to a story rather than to a subject. And you may simply not be the person who wants to make any of the three surviving films, which is a legitimate reason and doesn't require a grander one.
The cost that is not a reason: the teaser is shot. The treatment is written. Someone's cousin let you into the tower on a Sunday. Sunk work demonstrates that you have spent resources, not that a replacement film is warranted, and the more expensive the original project was, the harder that sentence is to act on.
Test whether the new question genuinely changes the material plan
Under the old premise the kit was: the tower, the face, the hands, the town beneath, a sunrise, a drone, and a keeper saying something about constancy.
Under the narrower question it becomes: a bench, a lathe, the log itself, brass, an oil can, the particular mess of a workshop that does this work, and someone's hands doing something small and exact. That's a different film in a different room with a different shooting ratio.
Under the myth inquiry it becomes: microfilm readers, ledgers, meeting minutes, the guidebook page, a plaque in a wall, a local history room, and possibly no clock at all. You'd be making an archive film, which is a form with its own demands and a much lower tolerance for padding.
Under the clock-as-record route it follows whatever the record names. A foundry mark. A supplier. A wage book. You don't know yet.
And under stopping, it follows nothing, which is the point.
Two tests make this concrete. First, does the new question require different access — different people, different permissions, different archives — and are those available or merely hoped for? "The workshop is open to us in June" is available. "We'll find someone who remembers" is a hope. Second, could you describe the new film without using the old shot list? A new heading over the old causal sequence is not a reframe. If the revised version still runs tower, hands, town, keeper, sunrise, with new narration laid over the top, you have changed the label on the box.
And a caution about the people. The correction may embarrass someone who repeated the old story in good faith. It is entirely possible to use a participant twice — first as the face of a myth, then as the person delivering a graceful account of how the myth was wrong. Asking a clock keeper to perform contrition for a story he inherited is a second way of getting your film out of him, and it's worth noticing when a shot list starts to require it.
Correct the active pitch, not just the research note
Now go find the sentence. It is not only in the research folder. It's in the logline, the first paragraph of the treatment, the temp narration over the teaser, the one-pager, the funding rationale, the website you built in a weekend, and the email you sent a commissioning editor in March. Each carries a different remedy, and the one that matters is whichever a funder has already read.
Then follow the dependency. Ask what else in the pitch was true only because the premise was. Loglines usually have a "why now" hook, and this kind of premise supplies it. Funding rationales often have a line about what the film reveals about a place or a temperament. If the film was going to reveal that continuity is possible, the rationale dies with the claim, and patching the logline leaves the body of the pitch quietly resting on nothing.
Tell the collaborators who need to know: the co-producer, the shooter who gave you a weekend, the editor who cut the teaser, the funder, the commissioning editor. What they need is the correction and what it changes. They do not need your research file. Keep a source trail for yourself — which document the finding came from, when it was made, what it establishes and what it doesn't, the date you reached the conclusion — because in eight months someone will ask why you dropped the clock film, and if you go the reporting route you'll want to know exactly where you left off.
One thing to refuse: don't leave the old premise up front as a hook whose correction lands later. A film genuinely about how a claim travelled is allowed to open with the claim, because the claim is its subject, and it will have reporting behind it. A pitch that opens with a fact it has already abandoned is a different move. It's dishonest by structure, and it teaches everyone you send it to that your stated premise is a device.
The decision
Write the sentence that failed, on one line, in its claim form. Wherever the record contradicts it, the claim comes out of every document where it currently appears as fact — the logline, the treatment, the one-pager, the emails already sent. That step is not one of the positions. It is done in all of them, including the one where you stop.
What follows is a choice of four, and you should be able to say which.
If a genuinely supported narrower question survives, and you want to make that film, and the access is available rather than hoped for, pitch it as a new project with its own shape. If what you now want to know is how the false account took hold, treat it as an unfunded reporting project until it has material; do not pitch a film you cannot yet source. If the records themselves are what you want — and you don't yet know what they support — give yourself a bounded reading-and-access period: a set number of weeks, a named archive, a written note on what you found and what it names. Treat that as unfunded work too, and let it earn a pitch only when the record has handed you something you can point a camera at. If none of those is true — or none of them turns out to be a film you want to make — stop. The claim has still come out of every document it was in, and that part was never optional.
Stopping doesn't get to be the noble option and it doesn't have to be the shameful one. The clock is still in the tower. The workshop is still open on Tuesdays. The entry is still four lines long, in a book nobody wrote for you, and it will say the same thing whenever you come back to it.
Frequently asked questions
What is the first move when evidence contradicts a film's premise?
Write the failed claim as a plain proposition, then inspect the record: who made it, why, what terminology it uses, what period it covers, and whether it actually reaches the claim. Do not announce disproof before that inspection, and do not manufacture a dispute if the record settles the question.
Why can a maintenance log be strong evidence but poor confirmation?
Maintenance logs exist to schedule work and account for money, so they are excellent evidence of the events they record and poor evidence of omitted events. A single entry can falsify a strong continuity claim, while quiet pages prove nothing. A narrowed record can still be decisive if the entry falls inside the period the claim covers.
What is the difference between reframing and stopping?
Reframing means choosing among honest routes: a narrower supported question, an inquiry into how the account developed, a different evidenced relationship to the records, or stopping. Each remaining route requires new reporting and access with uncertain returns; sunk work such as a shot teaser or written treatment does not warrant a replacement film.
How can you tell if a new question is a genuine reframe?
Ask what the camera follows and whether the new question requires different access that is available rather than merely hoped for. Can you describe the new film without using the old shot list? A new heading over the old causal sequence is not a reframe.
What must happen to an active pitch after the premise fails?
The failed claim comes out of every document where it appears as fact, including the logline, treatment, one-pager, and emails already sent, even if the project stops. Tell collaborators the correction and what it changes. Do not open with an abandoned premise as a hook whose correction lands later unless the film is genuinely about how that claim travelled and has reporting behind it.