Sources

Every claim in this course, and how much evidence it actually has.

Every source behind this course, with an honest note on how far each one actually reaches.

Knowledge

• Loewenstein, G. (1994). "The Psychology of Curiosity: A Review and Reinterpretation." Psychological Bulletin, 116(1), 75-98. The foundational paper. Curiosity = a felt gap between what you know and what you want to know. Use for: the core mechanism under every hook. Dense but primary; the whole course rests on it.
• Psychology Fanatic, plain-language summary of Information Gap Theory Gentle on-ramp to Loewenstein. Use for: a first read before the paper.
• Berger, J. & Milkman, K. (2012). "What Makes Online Content Viral?" Journal of Marketing Research, 49(2), 192-205. Analyzed every NYT article over 3 months. High-arousal emotions (awe, anger, anxiety) spread; low-arousal (sadness) does not. Use for: why emotional charge, not just topic, drives attention and sharing.
• Zeigarnik, B. (1927), via a 2025 meta-analysis review Open loops: unfinished tasks keep a low tension running. Use for: how a hook stretches curiosity past the first second. NOTE the honest caveat, the memory boost mostly failed to replicate; the pull to resume is what survives.
• Schultz / dopamine & reward prediction error, NIH review Dopamine fires on anticipation and peaks under uncertainty ("maybe"), not on the reward itself. Use for: why an uncertain payoff ("maybe this changes everything") is more magnetic than a guaranteed one.

Self-relevance: the targeting lever ("make it about them")

• Cherry, C. (1953), the "cocktail party effect" In a noisy room we filter almost everything out, but our own name cuts through instantly, even from a conversation we weren't listening to. Self-relevant information gets automatic attention priority. Use for: why an "if you're [X]" call-out stops the right scroller (it's like saying their name).
• Rogers, Kuiper & Kirker (1977), the self-reference effect Information encoded in reference to the self is recalled best (deeper processing). Use for: self-relevance doesn't just grab attention, it makes the hook stick.

Delivery channels: title (on-screen) vs verbal (spoken) hooks

_Note on trust: these are industry/vendor figures, not peer-reviewed. They point the same direction and come from named sources, but treat exact percentages as directional, not gospel, especially the older muted-viewing numbers._

• Digiday (2016): "85% of Facebook video is watched without sound" and Digiday: "75% of people watch mobile videos on mute". Publisher-reported, dated, but widely replicated. Use for: why a muted-readable TITLE hook (on-screen text) is not optional on video, a big share never hear the audio.
• Verizon Media & Publicis Media (2023), captions boost average watch time ~12%; viewers reported ~80% more likely to finish a video when captions/on-screen text are present. Use for: on-screen text earns real watch-time, not just accessibility.
• TikTok "3-second rule" round-up (Teleprompter, 2026): ~71% of viewers decide within the first 3 seconds; TikTok for Business found 63% of highest click-through videos hook within 3 seconds; scroll decision ~1-1.5s on TikTok, slightly slower on Reels (more visual). The 1-second hold is "decided by the first frame, before you speak." Use for: verbal-hook length (one breath, ~5-12 words) and why the FIRST FRAME/title does the earliest work.

Flavor mechanisms: why each shape works (Lesson 07)

_All verified against the papers' own abstracts/records. Trust level noted per entry. The blanket caveat we teach: none of these studied a person scrolling a feed. The machinery travels; the sizes and rankings do not._

• Kang, M. J. et al. (2009). "The Wick in the Candle of Learning: Epistemic Curiosity Activates Reward Circuitry and Enhances Memory." Psychological Science, 20(8), 963-973. 40 trivia questions; people rated confidence + curiosity before seeing answers. Inverted-U: curiosity peaked when confidence ≈ 0.50. The half-knowing point. Caudate (reward anticipation) activated while reading the question. High curiosity → better recall of surprising answers 1-2 weeks later. Use for: the curiosity gap, and as the measured version of our "sweet spot." Best single source in the whole flavor deck.
• Kivetz, R., Urminsky, O. & Zheng, Y. (2006). "The Goal-Gradient Hypothesis Resurrected." Journal of Marketing Research, 43(1), 39-58. Café "buy 10, get 1 free" cards: customers purchased faster as they neared the reward. Illusionary progress worked too, a 12-stamp card with 2 pre-filled beat a blank 10-stamp card. Use for: the process gap and the numbered promise (a visible finish line converts watching into progress). Caveat: purchases and effort, not video watching.
• Hart, W. et al. (2009). "Feeling Validated Versus Being Correct: A Meta-Analysis of Selective Exposure to Information." Psychological Bulletin, 135(4), 555-588. Moderate preference for congenial over uncongenial information, d = 0.36. The bias flips when the uncongenial info helps a current goal. Use for: the validation rung of the arousal ladder, and for the honesty caveat that validation is weaker than the internet claims and loses to usefulness.
• Fiorillo, C. D., Tobler, P. N. & Schultz, W. (2003). "Discrete Coding of Reward Probability and Uncertainty by Dopamine Neurons." Science, 299, 1898-1902. TWO separate dopamine signals: the phasic burst (prediction error) and a previously unobserved sustained ramp that grows until the expected reward time and covaries with uncertainty, maximal at P = 0.5. Use for: the physical shape of The Maybe (a rising line through the wait, not a spike), and specifically the bet. Caveat: monkeys, juice.
• Aronson, E., Willerman, B. & Floyd, J. (1966). "The effect of a pratfall on increasing interpersonal attractiveness." Psychonomic Science, 4, 227-228. A quiz contestant spills coffee. When he had answered 92% correctly the blunder made him more likeable; when he was mediocre (30%) the same blunder made him less likeable. Use for: the scar and honest uncertainty. And for the critical condition: a scar only builds trust if competence is already visible. Small 1966 lab study; directional.
• Hovland, C. & Weiss, W. (1951). "The Influence of Source Credibility on Communication Effectiveness." Public Opinion Quarterly, 15, 635-650. Identical message, different attributed source → more opinion change from the credible one; people discounted untrustworthy sources at the time of exposure. Also the origin of the sleeper effect: weeks later the advantage faded as people forgot who said it. Use for: credibility flavors generally, and for the point that credibility buys the first moment, exactly what a hook needs and no more.
• Loewenstein (1994), already listed above under Knowledge, also supplies two laws we apply to every gap flavor: curiosity requires some prior knowledge (you can't want what you can't frame), and small gaps produce high curiosity while large gaps produce low. NOTE: I could not verify the commonly-cited "five situational determinants" list (posing a question / unfinished sequence / violated expectation / someone else has the info / forgotten knowledge) against a primary source, so the lesson does not cite it as a list. What the lesson does use, and what the summaries do support, is his account of violated expectations causing arousal and triggering a search for an interpretation. The mechanism under the surprising-origin and emotional-contradiction gaps.

Flavors & field taxonomies (Lesson 07)

• videodatabase.org, "browse by hook strategy": seven hook strategies (H.01 Curiosity & Open Loops, H.02 Relatability & Contrast, H.03 Transformation & Payoff, H.04 Credibility, H.05 Secrets & Shortcuts, H.06 Opinions & Polarization, H.07 Identity & Specificity), each with creator lists and "popular concept" counts. Trust level: field observation, not research. It is a catalogue of visible surface patterns tagged across videos. Two hard caveats we teach alongside it: (1) the counts ("Breakdown 162," "Hot Take 147") measure how often creators use a thing in that crawl, popularity, never effectiveness; nothing there reports performance. (2) The sample has a visible tilt (several golf accounts recur across four strategies, alongside brands like Duolingo and Jacquemus) and is mostly English-language, so the rankings may not travel to Indonesian even though the mechanisms will. Also note the site's own line: videos use multiple strategies at once, these were never exclusive categories. Use for: a menu of shapes when stuck, and as the source of three genuine additions to our model, the process gap (H.03), the validation rung of the arousal ladder (H.06), and credibility as the deliberate bridge (H.04, flavors: receipt / tenure / scar / witness). Do NOT use as: a model of cause. It cannot see The Maybe, which has no phrase tell.