Matthew Dicks tells two anecdotes about being in movie theaters where the power went out mid-film.1
In the first theater, the screen goes black halfway through the third act. The audience erupts immediately — they are angry, they want the lights back on, they want their movie back. The theater staff scrambles to restore power. The audience members refuse refunds and wait in the dark for the projector to come back. When it does, they cheer. They watch the rest of the film. They leave talking about how they almost missed the ending.
In the second theater, the screen goes black at roughly the same point. The audience reacts — but the reaction is different. People shrug. People stretch. A few mutter that the movie was kind of slow anyway. Most of the audience picks up their bags and leaves. The theater staff offer to restore power; some audience-members accept refunds and leave; others simply leave without the refund. By the time the projector comes back, the theater is half-empty. The remaining audience watches the rest without much investment.
Dicks's diagnostic: the difference between the two movies was not whether they were good or bad films in the abstract; it was whether the audience cared about what happened next. The first film had built an audience that cared. The second film had built an audience that did not.
The same diagnostic applies to your story. If the power went out in the middle of your story, would the audience be angry? Would they ask you to finish later? Or would they shrug and check their phones? The answer separates story from anecdote. If the audience would care, you have a story. If the audience would shrug, you have an anecdote — possibly entertaining, possibly not, but missing the load-bearing element that makes a story land.
The test is a thought-experiment diagnostic for assessing whether a draft has the structural properties that produce audience investment in continuation.
1. The test operates at any point in the story. You can apply it at the end of the first paragraph, at the midpoint, at the climax — anywhere. If the power went out right now, would the audience care? The test is not a single end-point assessment but a continuous gauge you can apply across the story's runtime.
2. The test is binary at each point. The audience either cares or they do not. The test does not measure degree. This is part of its diagnostic value — most amateur storytellers conflate the audience is listening politely with the audience cares. The power-out test forces the distinction. If the lights cut and the audience would politely thank you and leave, they do not care. Only if they would actively want continuation do they care.
3. The test measures specific cognitive states. Caring in this test means the audience has formed a prediction-investment in the story's continuation. They have a hypothesis about what is coming. They want to find out if their hypothesis is right. They have invested attention in a specific question and they want the question answered. Stories produce this state; anecdotes do not.
4. The test is failure-diagnostic, not improvement-prescriptive. Failing the test tells you the story is not working but does not tell you why. You need other Dicks frameworks (Big Three, BABC, Mount Rushmore, four tools) to identify the specific structural failure. The power-out test is the alarm; the structural frameworks are the diagnosis.
The cognitive mechanism behind caring is prediction-investment. The audience has been listening; the listening has produced a model of what is happening in the story; the model has generated a prediction about what comes next; the audience wants to find out whether the prediction is right.
Prediction-investment requires three structural elements be present in the story so far.
Element A — stakes have been established. The audience knows what is at risk for the protagonist. Something matters. The protagonist could lose something, gain something, become something, fail to become something. Without stakes, the audience has no reason to care about what happens — the outcomes are interchangeable.
Element B — a specific question has been opened. The audience is tracking a specific unanswered question about the protagonist's situation. Will she tell him the truth? Will he survive? Will the project work? Will they reconcile? The question is precise enough that the audience can hold it. The story has narrowed the audience's attention to a specific unknown.
Element C — the audience has formed a hypothesis. Based on the established stakes and the opened question, the audience has formed an expectation about what is likely to happen. The hypothesis may or may not be correct — that is the storyteller's choice. But the hypothesis-formation is what produces the wanting to know. Audiences who have not formed a hypothesis are not invested in the answer.
When the power goes out, the audience that cares is the audience that has all three elements active. They have stakes; they have a question; they have a hypothesis. They want the question answered to find out whether the hypothesis is right. Cutting the lights at this point cuts them off from a payoff they had been actively constructing in their head.
The audience that shrugs has fewer than all three elements. Maybe stakes were never established (the audience never knew what was at risk). Maybe no specific question was opened (the audience was hearing events but not tracking an unknown). Maybe no hypothesis formed (the audience had no expectation about what would happen). Without prediction-investment, the lights going out is just the story stopping; nothing was being constructed in their head, so nothing is being cut off.
The power-out test operationalizes Dicks's story-vs-anecdote distinction at the audience-cognition level.
Anecdotes are sequences of interesting events without a load-bearing question. The audience hears the events, registers them as interesting or not, but does not form prediction-investment because no specific question has been opened. I went to a party and the strangest thing happened — followed by a description of the strange thing — is anecdote. The audience receives the strange thing and is mildly entertained. If the lights went out before the strange thing was described, the audience would shrug. They had no hypothesis about what the strange thing was; the strange thing was just coming; without it, there is nothing to be invested in.
Stories are sequences of events organized around a load-bearing question with stakes. The audience hears the events as evidence about how the question will resolve. I went to a party and saw the man who broke my heart in college standing across the room — opens a question (will she speak to him?) with stakes (the unresolved past, the present relationship she is in). The audience forms hypotheses (she will speak; she will avoid; she will pretend not to recognize him). They want to find out which. If the lights went out, the audience would care.
The two forms can use identical content and produce opposite power-out test results. The difference is whether the storytelling has organized the content around a load-bearing question. The same party, the same man across the room, can produce either an anecdote (a description of an awkward run-in) or a story (a question about what the protagonist did with the moment). The structural choices made by the storyteller — what to open, what to leave open, where to land the change — determine which.
The test is the diagnostic complement to the five-second moment of change. The five-second moment framework tells the storyteller what to land on (the moment of internal change); the power-out test tells the storyteller whether the audience is invested in the landing. Together they constitute the storyteller's two structural-diagnostic tools — one for the destination, one for the journey toward it.
It pairs with the no-one-wants-to-hear-anything mindset: the mindset is the storyteller's attentional default (assume the audience is unimpressed); the power-out test is the storyteller's verification (test whether you have nonetheless earned their investment). Mindset operates during drafting; test operates during revision.
It connects to elephant stakes: the elephant-stakes opening is one of the most reliable techniques for ensuring the power-out test passes early. Establishing stakes immediately gives the audience something to invest in from the first thirty seconds; without it, the test may fail in the first paragraph regardless of what comes later.
It links to the Big Three framework: the Big Three (find end / find beginning / master middle) is a structural workflow for constructing a story that passes the power-out test. Find-end produces stakes (the audience knows where the change is going); find-beginning produces a question (the audience knows what is at risk); master-middle produces the prediction-investment trajectory.
It contrasts with audience-happy-I'm-still-talking platonic ideal: the platonic ideal is the storyteller's positive-aspiration (the audience is glad you have not stopped); the power-out test is the negative-elimination (the audience would notice if you stopped). The two operate symmetrically — passing the negative elimination is the floor; reaching the positive aspiration is the ceiling.
Imagine the same biographical event handled by two different storytellers.
Storyteller A's version (anecdote): I was at my college reunion last weekend and ran into my freshman roommate. Haven't seen him in fifteen years. He looked completely different — gained a lot of weight, lost some hair, was wearing this terrible sweater. We had this awkward conversation about our jobs and our kids. Then he told me he had just gone through a divorce and was barely holding it together. I felt bad for him. We exchanged numbers but I don't think we'll actually call.
Power-out test applied at the midpoint (after terrible sweater): would the audience care if the story stopped? No. No stakes established (this is description, not arc); no question opened (the audience is not tracking any specific unknown); no hypothesis formed (there is no expected next-beat). The audience would shrug.
Storyteller B's version (story): I was at my college reunion last weekend when I saw my freshman roommate across the room. We had not spoken in fifteen years because of a fight in our senior year I still felt guilty about — a fight that ended our friendship. He had not seen me yet. I had a choice. I could pretend I had not seen him, walk to the bar, leave the reunion entirely. Or I could walk over and try to apologize for something I had done at twenty-one that he probably did not even remember.
Power-out test applied at the midpoint (after try to apologize for something I had done at twenty-one that he probably did not even remember): would the audience care if the story stopped? Yes. Stakes are established (the speaker's old guilt, the friendship's possibility of repair); a specific question is opened (will the speaker walk over?); a hypothesis is forming (some audience-members are betting on walk-over, some on avoid). If the lights cut now, the audience would want continuation. They would care.
Same biographical material. Same college reunion. Same former roommate. The structural choices — what to open, what to make load-bearing, what stakes to surface — are what produce or fail the power-out test. Anecdotes can be made into stories with the same content if the storyteller restructures around a load-bearing question.
You have a draft. You have read it multiple times and feel it is fine but suspect it is not landing as hard as it should. You want to apply the power-out test systematically.
It is Saturday morning, 8:47am. You're at the dining table with the printed draft and a yellow marker. You read the draft aloud. After each paragraph, you pause and ask: if the power went out right now, would the audience be angry, or would they shrug? You mark each paragraph with a Y (audience would care about continuation) or an S (audience would shrug).
The first paragraph gets an S. No stakes established yet. You make a note.
The second paragraph gets an S. Still no question opened. You make a note.
The third paragraph gets a Y. The protagonist has made a choice that creates a question (will the choice work?). You make a note.
The fourth and fifth paragraphs get Y. The question is still open; the audience is invested.
The sixth paragraph gets an S. The question got answered in the fifth paragraph, and the sixth paragraph is wrap-up that does not open a new question. The audience would now shrug if the lights went out — they have the answer they were waiting for.
You step back and look at the marked-up draft. The Y-zone is paragraphs 3 through 5. The S-zones are paragraphs 1-2 (no question opened yet) and paragraph 6 (question already answered).
The diagnosis: the story works but the structural shape is wrong. The Y-zone is too narrow. You need to compress paragraphs 1-2 (you spent too much real estate before opening the question; the audience drifted before getting invested). You may want to add a new question in paragraph 6 (or cut paragraph 6 entirely, ending on paragraph 5 where the question's answer is the landing).
You restructure. You compress the opening two paragraphs to half their original length. You cut the sixth paragraph entirely. The Y-zone now extends from paragraph 2 through the end. The S-zone is only the brief opening.
You re-read aloud and re-apply the test. The new draft passes the power-out test from paragraph 2 onward — meaning if the lights went out at any point from paragraph 2 to the end, the audience would care.
The pattern is: mark each paragraph Y or S, identify the Y-zone's start and end, restructure to maximize Y-zone length (compress pre-Y, cut post-Y, or relocate material so Y-zone extends earlier and later).
Audience members compliment specific lines but not the piece. Audiences who heard an anecdote (rather than a story) compliment what was interesting — that line about your sister was great — but they do not compliment the piece as a whole because the piece as a whole did not produce an integrated emotional response. Stories produce piece-level compliments; anecdotes produce line-level compliments.
You feel obligated to tell what happened next when asked. After an anecdote, listeners sometimes ask and then what happened? The storyteller experiences this as flattery — they cared about the continuation! — but the question is usually evidence that the listener registered the storytelling stopped without resolving anything they could close. The storyteller's anecdote was incomplete from the audience's perspective; the listener is asking for the resolution the storyteller never built in.
You cannot describe the load-bearing question your story answers. If you cannot answer what question does your story answer?, the story may not have a load-bearing question — in which case it is an anecdote, and the power-out test will fail at multiple points.
Your story's audience response is polite rather than active. Polite applause, polite that was nice, polite changing of subject — these are the marks of anecdote reception. Stories produce active responses: leaning forward during the listening, sustained eye contact, follow-up questions about the protagonist's experience, what happened to him afterward questions about the protagonist as a person rather than about the storyteller's craft.
You can move the order of paragraphs without changing the piece's effect. If your paragraphs are interchangeable, you do not have a story — you have a description with no temporal arc. Stories are structurally fragile; reordering paragraphs breaks the prediction-investment trajectory. Anecdotes are structurally robust to reordering because no trajectory was being built.
Dicks across both interviews holds the power-out test as one of the most useful diagnostic tools in his teaching — he describes returning to it across drafts, across performances, and across coaching sessions with other storytellers.2 The framing is consistent across his interviews — same anecdotes, same test, same operational deployment.
Lewis treats the test in his synthesis as a Dicks-signature diagnostic — Lewis recommends the test to his audience of would-be storytellers as one of the most reliable self-assessment tools available.3 Lewis's version of the framing is faithful to Dicks's framing with no added embellishment.
Jasper Claus operates a structurally analogous test in his corporate filmmaking practice — he describes asking what is the audience predicting will happen at this moment throughout his draft-evaluation, which is the same cognitive operation as the power-out test under a different name.4 Claus's version is articulated in less compressed form because his medium (corporate film with research and editing budgets) permits more elaborate testing protocols.
Watch what the test measures: the prediction-investment that distinguishes engaged from disengaged audiences. Multiple other domains have evolved analogous diagnostic tests.
Game design / narrative-game pacing: Game designers face the same problem at extreme scale — players will quit a game at any moment if their prediction-investment lapses. The industry has developed analogous diagnostics, often called next-button tests (would the player click the next-button right now?) and quit-rate analyses (at which screens do most players abandon?). Look at storytellers and game designers side by side — they face structurally identical attention-management problems and have evolved structurally identical tests. Put them together: Dicks gives you the qualitative single-piece version; game design gives you the quantitative aggregate-data version. Neither alone covers the full ground: storytellers cannot easily measure quit-rates because their audiences are captive in the room; game designers cannot easily test individual artifacts because their data is aggregated. Together they give both granularity and scale.
Journalism / news article testing: Journalists working in the digital era face an analogous problem — will the reader scroll past the lede? The journalism tradition has developed equivalent diagnostics — the scroll-test, the inverted-pyramid evaluation. A lede that would not survive a scroll-decision at paragraph two needs restructuring. Put storytelling and journalism together and the power-out test turns out to be a general structural-diagnostic for any time-pressured communicative form. Storytellers and journalists are managing the same attention-economy from different positions — storytellers compete with phones in audience pockets; journalists compete with infinite alternative articles one click away. Both have evolved similar tests.
Stand-up comedy / set construction: Stand-up comedians face an analogous test — if I cut this section, would the audience notice anything missing? The discipline of cutting filler is structurally parallel to the power-out test applied to set-construction. Comedians who cannot cut maintain bloated sets that lose audiences; comedians who cut aggressively maintain tight sets that hold. Pairing with Dicks: the comedy tradition supplies the cut-discipline (what to remove); Dicks supplies the audience-cognition framework (what makes the removal noticeable). The two together yield a more complete craft framework than either alone.
The Sharpest Implication The power-out test implies that the dominant failure mode in amateur storytelling is the absence of structural conditions for prediction-investment, not the absence of interesting content. Most amateur storytellers think their problem is that their material is not unusual enough — they need a more dramatic event, a more exotic location, a more famous protagonist. The power-out test reveals the actual problem: their material is not organized around a load-bearing question that the audience can invest in. The fix is not better material; it is structural reorganization of the material they already have. Almost any biographical event can become a story if it is reorganized around a load-bearing question with stakes. Almost any biographical event remains an anecdote if it is not.
Generative Questions