Roth knows the beginning of every movie he writes. He knows the ending. And then the middle, how much surprise?1 The interviewer asks. Roth answers in a single sentence that anybody who's tried to write a long form will recognize instantly: Middle seems like a great big blob. I have no clue.2
Then he tells you what he does about it. I get me, you know, it's like, what's the next three scenes?3 He writes them down — one word each. Wedding. Shootout. Whatever.4 Five scenes in a row outlined as single words. That's it. That's the entire structural plan he allows himself for the middle. Not a beat sheet. Not a treatment. Five words on a page, naming the next five things that have to happen. The blob is crossed one stepping-stone at a time.
The middle-as-blob principle is a structured permission to not know the middle in advance. Roth refuses to plan the second act in detail because detailed advance-planning of the second act tends to produce scenes that execute the plan rather than scenes that discover what the story is. The one-word outline preserves enough forward motion to keep writing while preserving enough emptiness for discovery.
This is not pantsing (writing without any plan). It's not plotting (writing from a full outline). It's a third mode: minimum-viable-forward-horizon. The writer always knows what the next five scenes are, by single word. The writer rarely knows what comes after those five. This produces a rolling fog where the immediate path is visible and the longer path emerges from walking the immediate path.
The middle is a blob because the middle is where character meets consequence. Act one establishes the problem; act three delivers the cost. The middle is where the writer has to render the escalation — and escalation is precisely the thing that resists pre-planning, because honest escalation requires the protagonist to make choices the writer hasn't anticipated. A pre-planned middle is a middle where the protagonist is moved like a chess piece. A discovered middle is a middle where the protagonist's choices feel earned because the writer earned them too.
Roth's principle catches the failure mode where writers panic in the blob, retreat to over-planning, and produce middles that read as treadmills.
Why one word? Why not three sentences per scene, or a paragraph?
The one word is a horizon-marker, not a content-plan. It tells the writer something will happen here without telling the writer what will happen here or how it will happen. The wedding will be a wedding. What people say at the wedding, who arrives uninvited, what the bride does in the bathroom before the ceremony — all of that is discovered when the scene is written.
The five-scene horizon is long enough to commit to but short enough to revise. Three scenes ahead and the writer is just sketching the next twenty pages. Ten scenes ahead and the writer has accidentally written a treatment. Five is the goldilocks zone where forward motion sustains without rigidity.
The one-word constraint protects against premature articulation. A writer who has written "At the wedding, Daisy realizes she still loves Benjamin and runs out" has committed to an interpretation of the scene before writing it. A writer who has written "Wedding" has only committed to the event. The interpretation emerges in the writing.
The middle-as-blob principle pairs precisely with the page-one-every-day re-entry ritual. Page-one re-entry maintains the writer's relationship to the early material. The five-scene outline maintains the writer's relationship to the immediate future. Together they keep the writer always inside a living draft, oriented backward and forward by a small distance, free in the middle to discover.
The principle also makes the theme-as-rudder principle workable. When the writer hits scene three and isn't sure what it's doing, they fall back on theme — how does this apply to the theme of the movie? The theme governs the discovery. The one-word outline only says wedding; theme says what kind of wedding.
Roth is writing a screenplay for Scorsese about the mafia coming to New Orleans in 1890. He knows the beginning (Sicily, 1860, two rival mafia families). He knows the end (New Orleans, after the lynchings). The middle is a blob.
Watch what he does in the interview, walking through one discovered scene: he had a scene where he wanted vengeance on a rival Don. He didn't pre-plan how the vengeance would happen. He let himself find it. He decided: sneaking into an opera. Then he decided: while the opera is going, I have him slit his throat. Then — and this is the discovered detail that elevates the scene — the guy whose throat's being slit is a guy who danced himself as kind of a singer. So he's doing a baritone with the opera singer.5
If Roth had pre-planned this scene at the outline stage, he would not have found the baritone detail. The detail required him to be inside the scene, seeing the rival Don as a man with a voice. The blob-mode permitted the discovery. A treatment-mode would have specified "vengeance scene at opera" and the writer would have executed it competently and missed what makes it memorable.
Then Roth extends the discovery — the opera singer is the only person who saw him cut his throat6 — and the singer becomes a recurring instrumental character through the rest of the script. One discovered detail in one discovered scene seeded a structural element neither the beginning nor the ending pre-planned.
You're three weeks into a screenplay. You have the opening sequence written. You have the closing image in your head. You sit down Monday morning and your forty pages of act two feel like a swamp.
Don't outline the whole middle. You can't, and trying will produce scenes you'll have to rewrite. Instead, take five minutes and write five words on a Post-it note. The next five scenes. Funeral. Phone call. Bar. Confession. Drive. That's it. Stick the Post-it on the monitor.
Open the script. Go to where you left off. Write the funeral scene. Don't write a setup — start in the middle, find what's happening, let it unfold. When you reach a natural stopping point, scroll back to the Post-it. Phone call. Whose phone? Calling whom? You don't know yet. Start writing. The character picks up. They listen. They say something you weren't expecting them to say. Now you're inside the next scene.
By Friday you've written four of the five Post-it scenes. They've gone places you didn't anticipate on Monday. You realize the Bar scene isn't going to work where it was — the Confession needs to come first, because the funeral and the phone call produced something that has to be confessed before the bar scene can land. You move them. You add a new Post-it word at position five: Decision. The horizon rebuilds itself.
This is not improvisation. The five-scene horizon was real. It was just appropriately small.
You'll know you've abandoned the blob-mode discipline when the act-two scenes start to feel like checking boxes. The wedding happens because the outline said wedding. Nothing in the wedding scene surprises you, the writer. If the writer isn't surprised, the audience won't be either.
The opposite failure is more common: the writer panics in the blob, retreats to detailed planning, and produces a middle where every scene executes a pre-decided emotional beat. Symptom: the writer can describe the scene before writing it, and the writing adds nothing that the description didn't already contain.
The deepest failure is the aimless blob — the writer who took the middle is a blob as permission to wander without any horizon at all. The five-scene Post-it discipline is what distinguishes generative blob-mode from undisciplined drift. Without the Post-it, the writer is just lost. With the Post-it, the writer is exploring a known immediate neighborhood while the long horizon stays open.
The blob-mode requires a specific cognitive disposition: tolerance for not-knowing while still showing up to write. Writers with high anxiety often cannot sustain it — the not-knowing reads as failure rather than as discovery, and the writer either over-plans (kills the discovery) or stops writing (kills the project). Roth's six-decade career and his I get to write affective stance may be prerequisites for the blob-mode discipline, not consequences of it.
Tension with Karlsson's long-incubation-then-fast-write approach: Karlsson tolerates blob-mode but uses it for whole essays, not for the middle of a long-form. The forms may not transfer. Screenwriters can blob-mode their middles because the spine and the ending hold the work together. Essayists may need full advance-clarity because the form doesn't have a spine to hold.
Gawande's book-as-math approach7 sits diagonal to Roth's blob-mode. Gawande chunks his book into 2000 words/week and 30-minute blocks — quantitative forward motion, not horizon-marker forward motion. Both are mechanisms for crossing the swamp. Roth's mechanism is small horizon, trusting that walking will reveal the longer path. Gawande's mechanism is small unit, trusting that volume will accumulate into the long path.
The convergence: both reject the full advance plan and both reject the no plan at all. The split: Roth's unit is the next scene-sequence; Gawande's unit is the next 30 minutes regardless of what's written. Reading both reveals that long-form survival has at least two distinct anti-blob disciplines, and writers may need both at different scales — Roth-mode for the middle's structural horizon, Gawande-mode for the day's quantitative output.
Karlsson's body-state writing summons the next move differently again — not from horizon or quantity but from somatic attunement. Three different anti-blob protocols for three different practitioner temperaments.
The compound insight: discovery-mode writing requires high ambiguity tolerance + small-reward operant scaffolding + structural spine to hold the work together. None of those three frames alone tells the full story. Roth's blob-mode is the practical synthesis. Writers who only know the psychology layer think it's a personality trait. Writers who only know the behavioral layer think it's a discipline trick. Writers who only know the screenwriting layer don't realize their craft is doing all three at once.
The Sharpest Implication
If you keep failing in the middle of long projects, the diagnosis is probably not that you need a better outline. The diagnosis may be that you need less outline. A detailed treatment kills the discovery the middle requires. A five-word horizon preserves it. The writer who can sit with wedding, shootout, whatever and trust themselves to find what those scenes contain is doing the actual craft. The writer who needs to know the wedding's full emotional beat before writing it is fleeing the craft.
Generative Questions