Hook Psychology• Lesson 07• Flavor deck card• The Hook System• Glossary
Lesson 07 · the shapes

The Flavors

Each engine can be poured into more than one shape. Here are all of them, plus a verdict on the 7 strategies you see everywhere.

Knowing the four engines is not the same as being able to build a hook on a Tuesday. What sits between them is flavors, the specific shapes each engine can take. Curiosity gap, utility gap, withheld noun, the bet. This lesson teaches all of them.

Engine vs flavor

An engine is a force in the reader's head. A flavor is the shape you pour that force into.

Think of heat. Heat is one force. But you can boil, fry, grill, or steam with it. Same energy, four completely different dishes. If you only know “heat,” you stare at the stove. If you know the four methods, you cook. Engines let you diagnose. Flavors let you produce.

Two rules before the deck:

Rule 1: flavors don't stack the way engines do You want all four engines running at once. You usually want one flavor per engine, done cleanly. Two gap flavors mashed into one line reads as noise, not depth.
Rule 2: the flavor must match what you actually have A flavor is a promise about the kind of payoff coming. Pick the one your content can pay. This is hook-payoff match, at the flavor level.

Your 7 strategies: the verdict

If you have spent any time reading about hooks you have met a list like this one: the seven “hook strategies” catalogued by videodatabase.org. Short answer: they're real, they're useful, and they are not a system. Here's the honest breakdown.

They were built by looking at thousands of videos and tagging what the words on screen look like. So they're a catalogue of visible shapes. Our four engines are invisible forces. That difference explains everything that's right and wrong about the list.

What's right: every one of the seven maps cleanly onto machinery you already know. Nothing in that list is new physics. Watch:

H.01 Curiosity & Open LoopsGap + Open LoopTwo of our engines wearing one label. The gap makes them stop; the loop makes them stay.
H.02 Relatability & ContrastSelf-relevance + Arousal (recognition)“That moment when” = the private-behavior mirror. “Expectations vs reality” = an emotional-contradiction gap.
H.03 Transformation & PayoffGap (process) + Arousal (hope)Shows the destination, hides the road. This is a gap flavor I never taught you, it's in the deck below now.
H.04 CredibilityNot a pull, the bridgeThe most interesting one. It doesn't make a gap; it makes people believe a real answer sits behind yours. See below.
H.05 Secrets & ShortcutsGap (utility), dressed as access“Secret / insider / hidden” raises the felt value of the missing how-to. Same engine, higher price tag.
H.06 Opinions & PolarizationArousal (anger + validation)Two rungs of the arousal ladder at once: people who disagree get indignation, people who agree get validation. Both are high-arousal.
H.07 Identity & SpecificitySelf-relevance, exactly“For [people] who…” is Lesson 06 with a different name. Nothing to add.
Where the Seven Strategies Actually Live Grid mapping the seven field-observed hook strategies onto the four real engines plus self-relevance and credibility, showing credibility as a bridge rather than a pull and the maybe as invisible to a surface tagger. GAP H.01 · H.03 · H.05 curiosity, transformation, secrets AROUSAL H.01 · H.02 · H.06 loops, contrast, polarization OPEN LOOP H.01 curiosity & open loops THE MAYBE — absent from the list — no visible phrase tell SELF-RELEVANCE H.02 · H.07 contrast, identity & specificity CREDIBILITY H.04 the bridge — not a pull

What's wrong with it as a system: the seven sit at four different levels of the machine and are presented as if they're siblings. H.07 is a targeting lever. H.04 is a trust device. H.01 is two engines. H.03 and H.05 are single flavors. Using that list as your operating system is like a recipe book where “heat,” “salt,” “Tuesday,” and “lasagna” are the four chapters.

How to actually use it Diagnose with the engines. Unstick with the seven. When a hook flops, ask which engine is missing, the seven can't tell you that. When you're staring at a blank page, run the seven as a menu of shapes you might not have tried. Menu, not model.

What the seven taught me: credibility is the bridge

This is the part worth stealing. Back in Lesson 01 you learned the believable bridge: a surprising gap only pulls if the reader senses a real answer exists. Without it, surprise turns into confusion, not curiosity.

I taught you the bridge as a feeling. The database shows creators building it deliberately, thousands of times, and there's 75 years of research under it.

The founding study is old and simple. Hovland & Weiss (1951) gave people the exact same argument, changing only who supposedly said it. The version from a credible source moved opinions more; people discounted the identical message when it came from a source they didn't trust.1 Same words, different result. That's what a credibility line buys you.

The twist that makes this a hook tool That same study found the sleeper effect: weeks later, the advantage faded. People remembered the message and forgot who said it, and the credible and non-credible versions ended up in the same place.1 So credibility mostly buys you the first moment, which is exactly what a hook needs, and only that. Don't expect it to carry a whole relationship.

Four ways to build it:

The receiptcraftA number that proves it happened. “400 videos, 3 years, and only 6 that actually landed.” Why it works. A precise number is expensive to fake and easy to check, so it signals you counted. Vague size (“so much”) costs nothing to say, so it proves nothing. No study I'd cite for this specific move. Reasoning only.
The tenureinferredTime served in the thing. “I edited other people's content for 5 years before I dared post my own face.” Why it works. This is the expertise half of source credibility: a reason to believe you'd know. Hovland & Weiss tested expertise-plus-trust as a package, not tenure claims in a caption. Mechanism from Hovland & Weiss (1951); the hook use is untested.
The scarresearchA mistake you admit to. Trust from failure, not success. “I deleted 200 posts because of one comment. Don't be like me.” Why it works. The pratfall effect. Aronson, Willerman & Floyd (1966) played people tapes of a quiz contestant who then spilled coffee on himself. When the contestant had answered 92% correctly, the spill made people like him more. A blunder humanises someone who looks superhuman.2 The catch you must not miss. When the contestant was mediocre (30% correct), the same spill made people like him less.2 The scar only builds trust if competence is already visible. If your whole brand runs on vulnerability, the competence still has to show somewhere, otherwise the admission just reads as someone who can't do the thing. Aronson, Willerman & Floyd (1966), Psychonomic Science. Small lab study, tape recordings, 1966. Directionally solid, not a law.
The witnesscraftSomeone else's words or screen, not yours. “Someone DM'd me: ‘I read this 4 times before posting.’” Why it works. It moves the claim outside you, so it stops being a thing you want and starts being a thing that happened. Also doubles as self-relevance: the reader recognises themselves in the witness. Sits under “social proof,” which is popular but broad. No single study I'd hang this on.
1 Hovland, C. & Weiss, W. (1951), “The Influence of Source Credibility on Communication Effectiveness,” Public Opinion Quarterly 15, 635-650. 2 Aronson, E., Willerman, B. & Floyd, J. (1966), “The effect of a pratfall on increasing interpersonal attractiveness,” Psychonomic Science 4, 227-228.

Look at what the database attaches to the Credibility strategy: its top formats are Journey Documentation (101), Breakdown (86) and Vulnerable Monologue (69). Its own example list literally includes “mistakes made.” That's the scar. Which means a brand built on admitting things is running a credibility strategy, building bridges out of embarrassment, whether or not anyone ever named it.

Careful: credibility and the Maybe are not the same trust Credibility is trust in the source (“this person would know”). The Maybe is honesty about the outcome (“I still don't know if it works”). They're strongest together: “I've tried this 30 times, and I'm still not sure it'll work for you.” Credibility without a maybe slides into overclaim, and you already know the certainty penalty punishes that.

What the seven cannot see

Scan the list again. Gap: there. Arousal: there. Open loop: there. Self-relevance: there twice. The Maybe: missing entirely.

That's not because the Maybe doesn't work. It's because the Maybe isn't a phrase. Curiosity has tells you can tag: “wait until you see.” Identity has tells too, like “if you're a.” The Maybe has no tell, because it isn't in the wording. It's whether the outcome is genuinely undecided. Two identical sentences, one honest and one rigged, look the same to a tagger and feel completely different to a human.

Any catalogue built from surface patterns will be blind to the engines that live under the surface. That's the strongest argument for keeping the four engines as your operating system and treating the seven as what they are: a list of shapes that happen to be popular.

Two honesty notes about that data

1: Those numbers are popularity, not performance “Breakdown 162,” “Hot Take 147”, those are counts of videos using a thing in that database. Nowhere does it say those videos performed better. A strategy can be common because it's easy to film, not because it works. Never read a frequency count as an effectiveness score.
2: It's one crawl, with a shape Look at the creator lists: golf accounts show up under four different strategies, alongside brands like Duolingo and Jacquemus. That's a sample with its own tilt, mostly English-language. The mechanisms underneath are human and will travel to Indonesian; the exact rankings may not. And the site says it itself: videos use several strategies at once, so these categories were never meant to be exclusive boxes.

The flavor deck

Everything the map promised, now actually taught. For each flavor: what's missing for the reader, why it works, an example in your voice, and how it fails.

I've tagged how much evidence each one actually has, because some of these rest on real studies and some are just things creators do that make sense. You should be able to tell which is which.

research  A named study tests something close to this.
inferred  A study establishes the mechanism, but nobody tested it as a hook.
craft  Observed practice and reasoning only. No study. Treat as a hypothesis.

Gap flavors: five ways to make something feel missing

Two findings govern all of them, so learn these once.

Law 1: a gap needs ground to stand on Loewenstein's review makes the point plainly: you can't feel curious about something you know nothing about. Curiosity needs a little existing knowledge to sit next to.3 Every flavor below has to land inside a world the reader already half-knows.
Law 2: small gaps pull harder than big ones Same review: gap size predicts curiosity, and it's the small gaps that produce high curiosity, not the huge ones.3 Kang and colleagues (2009) put a number on it. They gave 40 trivia questions and asked people, before seeing any answer, how confident they were and how curious they felt. The result was an upside-down U: curiosity was low when people were sure they knew, low when they had no idea, and highest when confidence sat around 0.50, the half-knowing point.4

That is your sweet spot, measured. Give them half the answer.
3 Loewenstein, G. (1994), “The Psychology of Curiosity,” Psychological Bulletin 116(1). 4 Kang, M. J. et al. (2009), “The Wick in the Candle of Learning,” Psychological Science 20(8), 963-973.
Curiosity Peaks at Half-Knowing Inverted-U line chart showing curiosity rising from no confidence in the answer, peaking when confidence sits around 0.50, and falling as confidence approaches certainty. the half-knowing point NO IDEA ≈ 0.50 CONFIDENT FULLY SURE CURIOSITY →
1 · Curiosity gapresearchMissing: what. A fact, an outcome, a thing that happened. “I sent one DM last year that kept me up for three nights.” Why it works. In the Kang study, brain scans showed the reward-anticipation area lighting up while people read the question, before any answer arrived.4 So the pull isn't intellectual. Wanting to know is processed like wanting a snack. The unanswered question is itself the reward being anticipated. Bonus. When curiosity was high, people recalled the surprising answers better one to two weeks later.4 A gap doesn't just win the stop; it makes the payoff stick. Fails when: the missing thing is vague. “Sesuatu terjadi” sits at zero confidence, the wrong end of the U. No imaginable answer, no pull.
2 · Utility gapresearchMissing: how. A method they'd use today. “How to talk on camera without redoing 40 takes.” Why it works. This one runs on a different force from all the others. Not “I must know” but “I will need this.” Berger & Milkman found practical usefulness was one of the traits that predicted an article getting emailed on.5 Sharing something useful also makes the sharer look useful, so it travels in DMs. Honest limit. That study measured sharing, not stopping. It's good evidence this shape spreads. It is not evidence it grabs. Fails when: they already know the how, or don't want the outcome. That's a no-gap, and it's the most common way a useful post lands flat.
3 · Surprising-origin gapinferredMissing: the link. Cause and effect that shouldn't go together. “My views went up when I stopped making good content.” Why it works. Loewenstein describes what happens when an expectation is violated: it causes bodily arousal and starts a search for an interpretation.3 That's the exact feeling. Your brain can't file the sentence, so it starts hunting for the missing rule. This is the one flavor that makes a gap and arousal at the same time, which is why it feels so strong. And why the bridge matters here most. The search has to be able to succeed. If no believable rule is reachable, the hunt aborts, and an aborted hunt feels like confusion, not curiosity. Fails when: it's just random. Random is the same shape as surprising and none of the payoff.
4 · Emotional-contradiction gapinferredMissing: how two feelings can both be true. “My most-viewed post is also the one that embarrasses me most.” Why it works. Same expectation-violation machinery as flavor 3, aimed at feelings instead of causes. We carry a script for which events produce which emotions: win → proud, viral → happy. Break that mapping and the search starts. The quiet advantage. A factual gap can be closed somewhere else. They could look it up. A contradiction inside your head can only be closed by you. There is no other source. Fails when: the contradiction is manufactured. If you didn't actually feel both, there's no real resolution waiting, and the payoff turns out to be a pose.
5 · Process gapinferredMissing: the road. You show the destination and hide the route. (This is H.03, Transformation & Payoff.) “From not daring to send a voice note, to going live for an hour. Here's the four weeks.” Why it works, part 1. It satisfies Law 1 in one move. Showing the end state hands them the knowledge that makes the gap felt. Before they see where you landed, there's nothing to be curious about. Why it works, part 2. The goal gradient. Kivetz and colleagues (2006) watched café customers on a “buy 10, get 1 free” card and found they bought faster the closer they got to the free coffee.6 A visible finish line turns effort into progress. A visible destination does the same to watching. The sneaky bit. The same study found the illusion of progress works too: a 12-stamp card with two stamps already filled in beat a blank 10-stamp card, even though both needed 10 purchases.6 Applied honestly: show them how far along the road already is. Fails when: nobody wants your end state, or you never show the road. Destination with no route is a brag, not a gap.
5 Berger, J. & Milkman, K. (2012), “What Makes Online Content Viral?” Journal of Marketing Research 49(2). 6 Kivetz, R., Urminsky, O. & Zheng, Y. (2006), “The Goal-Gradient Hypothesis Resurrected,” Journal of Marketing Research 43(1), 39-58.
Five Ways to Open a Gap Tree diagram showing the five gap flavors — curiosity, utility, surprising-origin, emotional-contradiction, and process — each with what it withholds and its evidence tier. GAP five ways something feels missing Curiosity gap missing: what RESEARCH Utility gap missing: how RESEARCH Surprising-origin gap missing: the link INFERRED Emotional-contradiction gap missing: how both are true INFERRED Process gap missing: the road INFERRED LEGEND research — a named study inferred — mechanism, untested as a hook

Arousal flavors: the ladder, plus one

You know the parent finding from Lesson 02: Berger & Milkman analysed three months of New York Times articles and found arousal, not good or bad feeling, predicted sharing.5 Here's what that study says rung by rung, and where it goes quiet.

Aweresearch“I didn't know that was possible.” Why it works. Awe is what you feel when something is too big for the model you were carrying, so the model has to stretch. Berger & Milkman found awe-inducing articles were more likely to be emailed on.5 It's the rare rung that's high-arousal and pleasant, so it costs you nothing in trust.
Anger / indignationresearch“That's not right.” Why it works. Anger is an approach feeling: it makes you want to move toward the thing and do something about it. On a feed, the cheapest available action is to comment or share. Berger & Milkman found anger positively linked to virality.5 The cost: it recruits people who disagree with you. That inflates the numbers without building an audience.
Anxiety / fearresearch“Wait, am I doing this wrong?” Why it works. Threat narrows attention onto the threat and holds it there. Anxiety was also positively linked to sharing in the same study.5 My caution, not the study's: your audience is overthinking introverts who already run high on this. Past a point, anxiety makes people avoid rather than lean in, and the avoiding looks exactly like a scroll.
Excitement / hopeinferred“Maybe I could too.” Why it works. Approach motivation pointed at a possible future version of themselves. It's the fuel under every transformation hook, and it pairs naturally with the process gap. Consistent with the arousal finding; not singled out as its own rung in that study.
Recognitioninferred“This is so me.” Why it works. This isn't a named emotion in the virality research. It's our build: the self-reference effect (Lesson 06) plus a jolt. The jolt is an expectation violation about privacy, you thought that behavior was only yours, and a stranger just described it. Assembled from Rogers, Kuiper & Kirker (1977) plus the violation mechanism. Nobody has tested “recognition” as an arousal rung.
Validation (new)research“Finally someone said it.” Relief at hearing a belief they held quietly. “Consistency is overrated if you don't know what you want to say yet.” Why it works. Hart and colleagues (2009) pooled the research on what people choose to read and found a preference for information that agrees with them over information that challenges them.7 The paper's title says it best: feeling validated versus being correct. The honest size. That preference measured d = 0.36, which is moderate, not overwhelming. And the same analysis found the bias flips when the disagreeable information is actually useful for something they're trying to do.7 So validation is real, weaker than the internet thinks, and it loses to usefulness. Fails when: the belief was already said out loud everywhere. Validation needs the belief to have been unspoken, otherwise you're agreeing with a crowd, not with a person.
7 Hart, W. et al. (2009), “Feeling Validated Versus Being Correct: A Meta-Analysis of Selective Exposure to Information,” Psychological Bulletin 135(4), 555-588.
One rule holds across all six, and it's mine, not a study's: the heat has to be one you actually feel. Performed anger and borrowed awe read flat, because arousal is a body state, and you can't fake a body state in five words.
Six Rungs of Arousal Tree diagram showing the six arousal flavors — awe, anger, anxiety, excitement, recognition, and validation — each with its trigger phrase and evidence tier. AROUSAL six ways to fire up the body Awe “I didn’t know that was possible” RESEARCH Anger / indignation “that’s not right” RESEARCH Anxiety / fear “am I doing this wrong?” RESEARCH Excitement / hope “maybe I could too” INFERRED Recognition “this is so me” INFERRED Validation “finally someone said it” RESEARCH LEGEND research — a named study inferred — mechanism, untested as a hook

Open loop flavors: five ways to hold the tension

Say the caveat first. These sit on softer ground than the gap flavors. The parent is Zeigarnik (1927), and you already know the honest version from Lesson 03: the memory half of that effect largely failed to replicate. What survives is the pull to resume an unfinished thing. So treat everything here as a working model, not settled science.

1 · Withheld nouninferredName that a thing exists, refuse to say what it is. “One sentence from my mom made me stop posting for a year.” Why it works. Naming the category (a sentence, from my mother) while hiding the content gives the gap a shape. The reader can hold an empty slot in mind, which is what Law 1 requires. “Something happened” gives them no slot, so there's nothing to hold. Fails when: you hold it too long. The debt goes stale, and a stale debt gets written off.
2 · Numbered promiseinferredA count opens a tab that stays open until the count closes. “Three things I do before posting. The third one's the hardest.” Why it works. The goal gradient again.6 A number is a visible finish line, and people push harder the closer the finish line gets. It also converts vague watching into countable progress: two down, one to go. Practical. Say the number early. The pull comes from seeing the distance shrink, and they can't see it shrink until they know how long it is. Fails when: the number is padded. Item three has to be worth the wait, or you've taught them not to wait next time.
3 · The turn (“but wait”)inferredClose one loop, and the closing itself opens the next. “…and it worked. Then week four, everything reset to zero.” Why it works. This is reward prediction error from Lesson 04, used mid-piece. A confirmed prediction produces almost no dopamine. A violated one produces a spike. So a real reversal doesn't just add information, it re-fires the same signal your opening hook fired. Fails when: the turn is fake. A reversal with no consequence is a prediction that was never really violated, and the brain isn't fooled twice.
4 · Unit cliffhangercraftEvery slide or beat ends pointing at the next one. Built for carousels. Slide 3 ends: “and this is what I got wrong for two years →” Why it works. Format logic, not psychology. A carousel makes the reader re-decide at every swipe. So the tension has to be renewed at every decision point, not created once at the start. A video asks “keep watching?” continuously; a carousel asks it discretely, N times. Fails when: every slide teases and none deliver. Pay something on each unit, not only at the end.
5 · Loop stackingcraftOpen the next loop before closing the current one, so a debt is always outstanding. Why it works. It's the scaling trick. One loop holds attention for one span. Overlapping loops mean the pull never drops to zero, because a new debt starts before the old one is settled. Fails when: you stack more than you can pay. Three open loops and one answer is a broken promise, and it's the fastest way to lose someone permanently.
Five Ways to Hold a Loop Open Tree diagram showing the five open loop flavors — withheld noun, numbered promise, the turn, unit cliffhanger, and loop stacking — each with its mechanism and evidence tier. OPEN LOOP five ways to keep the debt outstanding Withheld noun names it exists, hides what INFERRED Numbered promise a count opens a tab INFERRED The turn (“but wait”) closing opens the next INFERRED Unit cliffhanger every slide points forward CRAFT Loop stacking next opens before current closes CRAFT LEGEND inferred — mechanism, untested as a hook craft — observed practice, no study

Maybe flavors: three honest uncertainties

First, an upgrade to what you learned in Lesson 04. Fiorillo, Tobler and Schultz (2003) recorded dopamine neurons directly and found two separate signals, not one.8 There's the quick burst when a reward turns out different from what was predicted, that's the prediction error you already know. But there's also a slow, sustained climb that builds through the whole wait, and that one tracks uncertainty. It is at its maximum when the odds are 50/50, and it shrinks as things become more certain in either direction.8

That's the physical shape of a maybe. Not a spike at the end. A rising line through the middle, biggest when the outcome is a genuine coin-flip. It's why the wait is the part that holds people.

1 · Honest uncertaintyresearchYou genuinely don't know if it transfers to them, and you say so. “This works for me. I'm not sure it'll work for you, and I'll tell you why.” Why it works, two engines at once. The uncertainty keeps the sustained ramp running,8 and admitting a limit is a pratfall: it humanises you.2 The same catch as the scar. The pratfall only helped the contestant who was already visibly competent.2 Doubt from someone who has shown nothing reads as “this person doesn't know.” Earn it first, then admit it. Fails when: you're actually sure. Fake doubt is a rigged tease with better manners.
2 · The betresearchYou stake something in public where you could really lose. “30 days posting every day. If nothing changes, I'll post the numbers as they are.” Why it works. This is the 50/50 finding used literally. A bet you might genuinely lose sits near the probability where that sustained climb is at its highest.8 A bet you've already won sits near certainty, where the climb flattens out. Same words, different curve. Honest about the source. Those were dopamine neurons in monkeys waiting for juice. The machinery is shared and the shape is striking, but nobody has run this on a person watching a Reel. Fails when: you can't lose. A bet you've already won is a brag with a countdown attached.
3 · The skeptic's turncraftStart as the doubter, let the outcome decide. “I don't believe in ‘wake up at 5am.’ I'm trying it for two weeks so I can laugh at it.” Why it works. It puts you on the reader's side of the table. The objection they were about to make is the one you just made, so you're not selling them anything, you're testing it with them. It also makes the outcome genuinely open in a way a believer's version never is. Fails when: the verdict is obvious from the first line. If they can guess the ending, the odds aren't 50/50 any more, and the climb never starts.
8 Fiorillo, C. D., Tobler, P. N. & Schultz, W. (2003), “Discrete Coding of Reward Probability and Uncertainty by Dopamine Neurons,” Science 299, 1898-1902.
Three Honest Uncertainties Tree diagram showing the three maybe flavors — honest uncertainty, the bet, and the skeptic's turn — each with its mechanism and evidence tier. THE MAYBE three ways the outcome stays open Honest uncertainty you genuinely don’t know if it transfers RESEARCH The bet you stake something you could lose RESEARCH The skeptic’s turn start as doubter, let outcome decide CRAFT LEGEND research — a named study craft — observed practice, no study

How much should you trust any of this?

Enough to build with. Not enough to argue from.

Look at what these studies actually did: 40 trivia questions in a lab. A coffee shop loyalty card. Tape recordings of a 1966 quiz show. Monkeys waiting for juice. New York Times articles from 2008 and 2009. Not one of them studied a person scrolling Reels at 1am.

What travels is the machinery, the half-knowing peak, the ramp under uncertainty, the humanising blunder. What does not travel is the sizes and the rankings. Nobody has measured whether the process gap beats the curiosity gap on Indonesian Reels, and anyone who tells you they have is selling something.

So: use the flavors to generate, use your own posts to decide. The research tells you which levers exist. Only your audience tells you which one to pull.

Next: the capstone Lesson 08 · Building One Cold. The build order that turns all of this into a hook you can produce on a Tuesday, plus a real 5.4M-view hook taken apart, including why it breaks two of my rules and works anyway.

Practice: name the shape

Do this before the capstone Take one idea you actually want to post. Write the hook three times, using three different gap flavors from the deck (e.g. utility, surprising-origin, process). Then add a credibility bridge by scar to whichever of the three feels weakest, the admitted mistake, not the win. Write all three below, say which flavor you'd ship, and why.
Source The seven hook strategies come from videodatabase.org's “browse by hook strategy” taxonomy. Treat it as a well-observed field catalogue of surface patterns, not a validated model of cause. The engines it maps onto are in Sources with their primary sources.