Talk Narrative and Structure
🗒️ Description
The decisions that settle what goes into a talk and in what order, made before any slide exists. This is the second wave of the research behind Conference Talk Craft. That note covers the opening, the clock and Q&A — what happens on the day. This one covers the design work upstream of the day. Eleven techniques, each researched against primary sources and graded for evidence.
The pattern from the first wave repeats: almost every technique here is sound practice, and most come with a justification that does not survive checking. I have kept the practice and replaced the reason each time, and said so below instead of quietly dropping the myth.
The siblings: Slides for Technical Talks for the visual layer, Delivery and Stage Operations for the day itself, Practising and Reviewing Talks for getting better across talks.
🧩 The techniques
Decide what the talk is for (design time, under an hour each)
| Technique | What it is | On stage |
|---|---|---|
| Throughline and my own why | One sentence of 15 words or fewer that every part of the talk exists to support, plus one sentence on why I am the person entitled to say it. Anything that serves neither is cut. Highest priority of the set: the opening, the cutting and the closing all depend on it. | 0.5 min |
| Point A to Point B (WIIFY, AFTERs) | Write down where the audience starts and what they will do, know or believe when they leave. Every block either moves them from A to B or goes to the appendix. A cutting filter that is distinct from the thesis. | 0.5–1 min |
| Talk archetypes | Choose the skeleton before writing: problem-solution, journey or migration, before-and-after, failure case study, tutorial, myth-busting. Each one implies a different opening, block order, type of evidence and ending. | 0 |
| Triad | At most three main points. If I have five, two are merged or cut. | 0 |
| One example carried through | One concrete running example, developed from motivation to punchline. When time runs short, cut the general case, never the example. Typically saves 2–4 minutes. | 0 |
| Audience level calibration | Declare the level as concrete prerequisites in the abstract, confirm it in one sentence on stage, pitch every block at it, and accept that some people will find it too easy. | 0.5 min |
| Curse-of-knowledge test | Once I know something I cannot reliably imagine not knowing it. The fix is not trying harder to empathise but testing a block on someone outside my specialism and cutting or explaining whatever they trip on. | 1–2 min of definitions |
Shape the argument
| Technique | What it is | On stage |
|---|---|---|
| SCQA and the Minto pyramid | Open with Situation, Complication and the Question it raises, then give the Answer as one governing idea supported by a few grouped, non-overlapping arguments. Strongest for decision-oriented and client talks; a dependable opening skeleton elsewhere. | ~1 min |
| Duarte sparkline and the STAR moment | Alternate between what is and what could be, and design exactly one “Something They’ll Always Remember” moment that makes the gap undeniable. Medium priority: thin evidence, but useful for persuasive talks and for clips. | 1–2 min for the STAR moment |
Keep the room oriented
| Technique | What it is | On stage |
|---|---|---|
| Signposting and progress cues | Announce the structure once, mark every major transition aloud, keep a small breadcrumb on the slide showing the current section. Whoever drifted for a minute can tell where the talk is and rejoin. High priority, and the best-evidenced item in this note. | ~1 min |
| Bookending, foreshadowing, backtracking | Close on the image or question I opened with, plant an element early and pay it off late, and restate where we are at each block boundary. Cheap; the bookend and the backtracking carry most of the value. | 1–1.5 min |
⚖️ What the evidence actually supports
Sound practice, wrong justification
| Practice | The usual reason | What the research says |
|---|---|---|
| Write a why-me sentence | ”You can’t inspire others unless you are passionate” | Chen, Yao & Kotha 2009: preparedness, not perceived passion, predicted venture funding. Justify the sentence as earned credibility, not as contagious enthusiasm. |
| Design one STAR moment | ”Vivid moments persuade” | The vividness effect is weakly supported, and seductive details cost learning (Sundararajan & Adesope 2020, g = −0.33). A STAR moment works only when it is itself the evidence for the main idea. |
| Cap at three points | ”Working memory holds three items” (or seven) | A short-term-memory limit misapplied to a 25-minute talk; capacity is nearer four chunks (Cowan 2001). Better support: impressions of a persuasive source rise up to three claims and fall at the fourth (Shu & Carlson 2014), plus simple time per point. |
| Plant and pay off | ”Open loops are remembered better” (Zeigarnik) | A 2025 meta-analysis found no memory advantage for unfinished tasks; only the tendency to resume them held up. |
| Close on the opening | ”The recency effect” | Recency is exactly what vanishes under a filled delay (Glanzer & Cunitz 1966). The defensible reasons are closure, reinterpretation of the opening, and re-entry. |
| Signpost | ”Attention drops after ten minutes” | Folklore (Wilson & Korn 2007). Lapses are short and frequent, which is the reason a re-entry point helps. |
| One running example | ”One example is all you need” | Transfer from a single case is weak (Gick & Holyoak 1983; Gentner, Loewenstein & Thompson 2003). Use the running example, a stated principle, and one short contrasting case. |
| Answer first (SCQA) | “Answer-first always wins” | Minto built the pyramid for written documents that busy readers skim. In a live talk the audience cannot skip ahead, and her own introduction framework is SCQ with the answer at the top of the pyramid. |
| Put myself in their shoes | ”Just empathise” | Simple perspective-taking does not remove the bias (Hinds 1999). Testing on real people does. |
| Myth-busting archetype | ”Debunking backfires by repeating the myth” | The backfire effect is now considered rare. The Debunking Handbook (2020) recommends stating the myth once, clearly flagged. |
Two more cautions that are not about justification. A call to action at the end is not mandatory: Anderson lists the ritual call to action among the clichés that can feel manipulative, and Point B is a designed change, not necessarily an instruction. And a mixed-level room cannot be fully served by careful layering: the expertise reversal effect says the same material is not optimal for novices and experts at once, so pick the level and say so.
False attributions to avoid
- “Curse of knowledge” was coined by Camerer, Loewenstein & Weber (1989), not by the Heath brothers who popularised it. The famous tapper numbers come from Newton’s unpublished 1990 dissertation, which contradicts itself: its table gives 3 hits in 120 tries (2.5%) against a predicted 50%, while the abstract says 2 of 150 and a later page says 3 of 150. I cite “2.5% (3 of 120) against 50%” and say it is one unreplicated dissertation.
- “Bookends” in Presentation Patterns means something else: distinct slides that supply agenda context throughout the talk. The closing that returns to the opening is storytelling and speechwriting craft, not a Presentation Patterns pattern. Foreshadowing and Backtracking are genuine entries there.
- Cold Open, Talklet and the other wave-one attributions are in Conference Talk Craft.
Where the evidence is strongest
- Signalling is controlled-study evidence (Mayer), with the usual caveat that it comes from short learner-paced lessons, not live talks.
- The curse-of-knowledge bias itself is robust across paradigms; only the famous numbers are shaky.
- Expertise reversal is a well-replicated controlled finding, which is why level calibration matters more than it sounds.
- Everything else here is practitioner consensus: widely agreed, cheap, and untested in a conference room.
🔧 Cheapest first moves
- Throughline plus Point B on one card, before opening the deck. Then mark every slide U, B, A or X (does it help them understand, believe, act — or none of these) in under ten seconds each, and move the X slides to an appendix. The count of X slides is the cut list.
- Ten-to-three: write every point I want to make on its own line (8–12 of them), force them into three clusters, send the rest to a backup pile.
- Explain-back: deliver one block to someone outside my specialism, then ask them to say it back without looking. What they cannot repeat is what to cut or define.
- Transition-only run: rehearse just the three block handovers, five times each, from 20 seconds before the end of one block to 20 seconds into the next.
📖 Further reading
- Conference Talk Craft — wave one: the opening, the clock, Q&A, demos, evidence grading
- Public Speaking — hub note for the whole cluster
- Slides for Technical Talks · Delivery and Stage Operations · Practising and Reviewing Talks — the rest of wave two
- TED Talks — Anderson on the idea as payload, and on call-to-action clichés
- The Art of Public Speaking — Carnegie’s structure of a speech, a century earlier
- What Carries Over Between Talks — my own record of how preparation actually goes
Template: knowledge_note_info