Protocol

This file records how the pool of sources for this topic was built. With it, the search can be repeated.

It follows Garousi 2019 for a multivocal literature review (MLR, defined in the evidence README) in a lighter form.

Garousi 2019 was written for software engineering. Its guidelines are applied here to presentations, a field with a large teaching and speaking practice beside its research. The formal pool predates the protocol, and this file says so.

1. Why grey literature (Garousi Table 4, seven questions)

# Question Answer for presentations
1 Complex subject, not solvable by the formal literature alone yes: speaker, room and format rules are mostly practice
2 Lack of volume, quality or consensus of evidence yes: few live-talk studies, bundled factors
3 Context matters yes: audience, length, room, field
4 Goal to check research against practical experience yes
5 Goal to challenge practice rules with research yes: 30 point, 10 minutes, 3 seconds
6 Synthesis useful to both communities yes
7 Many practitioner sources yes: every grey string returned speaker guides and workshops

Seven of seven are yes, so grey literature is included. Q7 (motion) and Q8 (colour) get the same seven answers: palettes, builds, demos and video style are mostly practice, and the experiments are few and small.

2. Questions

  • Q1 Does on-screen text that repeats the voice help or hurt, and which principles apply to a paced medium?
  • Q2 Does showing slides help, and which slide design helps?
  • Q3 What does the speaker add: gesture, face, enthusiasm?
  • Q4 What makes slides legible and accessible in a room?
  • Q5 How are talks built, and how do questions and discussion work?
  • Q6 What does practice recommend and adopt, and how does it compare with Q1 to Q5?
  • Q7 Motion in a talk: slide builds and transitions, animated diagrams, embedded video, explainer video in the style of mathematical animation, live demos and live coding, lecture-video design.
    • Its core: what changes when a speaker paces the motion.
  • Q8 Colour: colour coding as a signal, emotional design with pleasant colours, colour in data figures, colour-vision deficiency and safe palettes, light against dark when projected, colour harmony, and what practice advises.

3. Existing reviews (Guideline 2)

Id String, as run Results screened Included
E1 systematic review meta-analysis PowerPoint slides student learning outcomes 10 0
E2 meta-analysis multimedia learning principles review of reviews Mayer 10 2 (Noetel 2022, Cromley 2025)
E3 systematic review scientific conference presentations oral presentation slides effectiveness 9 1 (Blome 2017)
E4 meta-analysis instructor visibility lecture video learning outcomes 2022 2023 2024 10 3 (Polat 2022, Zhu 2024, Heidig 2024)
E5 scoping review slide design PowerPoint lectures higher education text density images 9 0 (Le Blanc 2025 came through G4)
E6 meta-analysis teacher gestures learning students instruction 9 0
E7 systematic review OR meta-analysis animation versus static slides presentation learning builds progressive disclosure 10 0 (animation reviews already held in the linked learning-science review)
E8 meta-analysis video lectures learning higher education systematic review lecture video design 9 1 (Noetel 2021)
E9 live coding versus static slides teaching programming study learning outcomes 9 2 (Shah 2023, Shah 2025)
E10 Live Coding: A Review of the Literature Selvaraj ITiCSE 2021 9 1 (Selvaraj 2021)
E11 systematic review color coding signaling multimedia learning color as cue meta-analysis 9 0 (Richter 2016, Schneider 2018 and Noetel 2022 already held)
E12 experiment sequential bullet points reveal versus all at once slides recall learning "build" animation PowerPoint study 10 2 (Ito 2026, Liu 2022)
E13 explainer video animation versus slides lecture learning engagement evidence whiteboard animation study 10 2 (Li 2023; Pichler 2023 as grey)

Engine for E, F and G strings: a general web search tool (one engine, US-based, returns a summary and links), run on 2026-10-28. E7 to E13 screened 66 results.

7. Selection

  • Include: a source that states a rule or gives evidence on Q1 to Q6, in English, a talk or lecture with slides as its unit.
  • Exclude: sources about posters or online meetings; pages that repeat another page; sources unreadable here. Video production is excluded as a craft (cameras, lighting) but included as design research (Guo 2014, Brame 2016).
  • For Q7 and Q8 also: colour-only studies of web pages, packaging or marketing (no listener, no slide); animation of pedagogical agents (covered under Q3 in the first round); sources that are vendor templates.
  • A grey source is scored (section 8); one under 10 is kept only as an example of practice.

8. Quality of grey sources (Guideline 11, Table 7)

  • 20 criteria, each 1, 0.5 or 0: authority of the producer (4), methodology (6), objectivity (4), date (1), position among related sources (1), novelty (2), impact (1), outlet type (1).
  • Impact is read here as 1 when a count or derived guides were seen, 0.5 for evident wide use or an owner's claim, 0 otherwise.
  • Outlet is 1 for books, standards and reports, 0.5 for owner guides and annual reports, 0 for blogs.
  • Books known from a catalogue page only are scored on what that page shows; sources.md says so.
  • Threshold 10 of 20 (Garousi): 19 of 39 grey sources are under it (12 of 24 before the Q7 and Q8 round, 7 of 15 in it).

9. Extraction and synthesis

  • Per source: producer, date, claim, number, evidence cited, limits.
  • Synthesis is narrative by question; each claim traces to a row of sources.md.

10. Limits

  • A second rater re-scored 5 of this topic's grey sources blind (method), before Q7 and Q8 were added: 4 of 5 decisions agree. The TEDx organiser guide scores 9.5 and 12.5, as the raters weighed an owner's account of its own format differently; it stays an example of practice. The sources added for Q7 and Q8 were scored by one rater only.
  • One search engine that personalises and changes; results of 2026-10-28 may not repeat.
  • Screening used titles and engine summaries; secondary numbers are marked "(secondary)".
  • Formal side: ten of the frame terms counted in the first round, the pool not screened from them. In the Q7 and Q8 round the counts come from ERIC and arXiv, not from OpenAlex, so they are not comparable with the first table.
  • Many Q7 and Q8 papers sit behind publisher paywalls; where only the abstract was read, sources.md says so, and numbers carried only by a search summary are marked "(secondary)".
  • Whether Richter 2016 and Schneider 2018 test colour as a moderator could not be settled: their full texts were not reachable, and only the abstracts and an archive description were read.

Q9 business presentations (added 2026-10-29)

Numbering continues the strands E (E14 on), F (F3 on) and G (G32 on).

Table 4 for Q9 (Garousi seven questions)

# Question Answer for business presentations
1 Complex subject, not solvable by the formal literature alone yes: slide conventions of firms, accelerators and sales methods are practice
2 Lack of volume, quality or consensus of evidence yes: investor decisions have a literature; slide form, memo against deck and executive briefings have almost none
3 Context matters yes: investor, buyer, executive and board decide on different things
4 Goal to check research against practical experience yes
5 Goal to challenge practice rules with research yes: passion, "answer first", "no slides"
6 Synthesis useful to both communities yes
7 Many practitioner sources yes: consulting guides, accelerator guides, deck-hosting data, sales training firms

Seven of seven are yes, so grey literature is included.

Q9 questions

  • Q9a Structure: answer first (pyramid principle, BLUF, action titles, the SCQ frame) against the academic build-up. What is measured, what is only practice?
  • Q9b Decision audiences: what investors, buyers and executives decide on; perceived passion and preparedness; the presenter against the content.
  • Q9c Formats: slide deck against written memo; the deck read without a speaker; how pitch decks are read.
  • Q9d Persuasion and attention: what the evidence on persuasion, storytelling and data storytelling supports in a business setting.
  • Q9e Practice: consulting conventions, sales methods, accelerator guides, design-firm guides.
  • Q9f Against: critiques of slides in organisations, and where business practice contradicts Q1 to Q8.

Engine and date

  • E, F and G strings: a general web search tool (one engine, US-based, returns a summary and links), run on 2026-10-29. A "screened" count is the number of result blocks returned; each block was read as a title plus the engine's summary.
  • Metadata and DOI checks: the public Crossref REST API (api.crossref.org/works), called with no identity and no mailto. Its answers confirm title, authors, journal, year and pages; they carry no abstract in most cases.
  • Pages opened one at a time. Eight opens failed (listed in section "Seen, not retrieved" of sources). No sign-in, no CAPTCHA, no bypass was tried.

Seed checks (not counted in the screening bound)

Each seed was looked up, then its page or abstract opened where open.

Seed Exists Says what was cited Read how
Clark 2008 yes, Venture Capital 10(3), 257-279 yes: 24 UK angels rated 3 pitches on 32 aspects; delivery factors weighed most; angels unaware abstract (IDEAS page)
Chen 2009 yes, AMJ 52(1), 199-214 yes, with a twist: preparedness, not passion, went with funding abstract (indexed copy)
Pollack 2012 yes, ETP 36(5), 915-939 yes: televised pitches, preparedness to legitimacy to funding abstract (Crossref record text)
Brooks 2014 yes, PNAS 111(12), 4427-4431 yes, and it adds attractiveness: helped men, not women abstract and full text (author copy, MIT DSpace)
Huang 2015 yes, ASQ 60(4), 634-670 partly: it is about investors' gut feel, not about presentations abstract (Crossref record text)
Kosslyn 2012 yes, already held yes: 140 slideshows, 137 rules; research slides no better than business slides overall abstract and results summary (open access page, first 100,000 characters)

Seed searches returned 59 result blocks (Chen two strings of 10, Clark 10, Brooks 10, Pollack 9, Huang 10).

E: existing reviews (Guideline 2)

Id String, as run Engine Screened Included
E14 systematic review entrepreneurial pitch investor decision making pitch presentation literature review web search 9 2 (Kalvapalle 2024, Ferrati 2021)
E15 meta-analysis narrative persuasion versus statistical evidence storytelling effect size web search 10 1 (Zebregs 2015); Shen, Braddock and Allen meta-analyses seen, not opened
E16 Kernbach Eppler Bresciani PowerPoint constraining qualities literature review organizations decision making web search 10 0 (wrong query: unrelated pages)
E17 Kernbach Bresciani Eppler "Slip-sliding-away" review literature constraining qualities PowerPoint Business and Professional Communication Quarterly web search 10 1 (Kernbach 2015)
E18 data storytelling narrative visualization review Segel Heer genres; Kosara Mackinlay storytelling next step visualization evidence web search 20 2 (Segel 2010, Kosara 2013)

E16 returned nothing useful; E17 is the same query done correctly (zero results meant the query was wrong).

F: formal

Id String, as run Engine Screened Included
F3 Clingingsmith Shane training aspiring entrepreneurs to pitch experienced investors field experiment results web search 9 1 (Clingingsmith 2018)
F4 Bernstein Korteweg Laws attracting early-stage investors randomized field experiment team traction web search 10 1 (Bernstein 2017)
F5 Kanze Huang Conley Higgins "We ask men to win and women not to lose" venture capitalists questions Academy of Management Journal web search 9 1 (Kanze 2018)
F6 experiment conclusion first versus conclusion last message order comprehension recall business writing inverted pyramid bottom line up front study web search 18 1 (Britt 2003; first seen through a summary, read in full on 2026-10-29)
F7 Yates Orlikowski PowerPoint presentation and its corollaries: how genres shape communicative action in organizations web search 20 1 (Yates 2007)
F8 experiment written narrative memo versus bullet slides executives decision quality comprehension study business PowerPoint web search 19 0 (opinion pieces only; no experiment found)

Also found, and not included, since only a title or a summary was read:

  • Allen and Preiss: a meta-analysis of narrative against statistics (1997);
  • Shen and colleagues: a meta-analysis of narrative in health messages (2015; 25 studies, N = 9,330);
  • Braddock and Dillard: a meta-analysis of narrative persuasion (2016);
  • Elsbach and Kramer: a study of Hollywood pitch meetings (2003);
  • Mitteness and colleagues: a study of angels and perceived passion (2012).

Cardon 2017 (full text, read 2026-10-29) and Yang 2021b (abstract reading) are included for the passion debate.

Forward and backward snowballing: not run (no citation index without an identity). Cardon 2017 and Yang 2021b came from the summary of the Chen 2009 string.

G: grey

Types and producers (Guideline 6):

Type Producers Why
Deck-hosting data DocSend (Dropbox) the only large count of how decks are read
Investor guides Sequoia, Y Combinator what founders are told to show
Consulting practice StrategyU, Minto International action titles, storyline, pyramid
Company practice Amazon letter, Tufte's page on it memo instead of slides
Sales methods Huthwaite (SPIN), Challenger what sellers present and ask
Military writing standard US Army correspondence regulation origin of BLUF
Id String, as run Engine Screened Opened Included
G32 DocSend pitch deck interest study how long investors spend reading pitch deck seconds web search 9 3 2 (DocSend 2022, DocSend 2026)
G33 Army Regulation 25-50 bottom line up front BLUF correspondence writing standard web search 10 2 0 (the regulation itself was not reachable)
G34 consulting slide conventions action title ghost deck storyline McKinsey Bain BCG how consultants write slides web search 10 1 1 (Millerd 2025); no firm's own document found
G35 Bryar Carr Working Backwards Amazon narrative six-page memo silent reading meeting PowerPoint banned web search 9 2 1 (Tufte 2022); Bezos 2018 re-found
G36 Sequoia Capital pitch deck template Y Combinator how to build your seed round pitch deck guide web search 9 2 1 (Sequoia 2019); the Y Combinator page returned a title only
G37 Rackham SPIN Selling research 35,000 sales calls Huthwaite; Challenger Sale Gartner research buyers web search 18 1 1 (Huthwaite 2026); Challenger figures stayed secondary

Added directly, not through a string: Minto 2026 (the author's site, opened to check the pyramid and SCQ claims).

Totals: E 59, F 85, G 65 = 209 result blocks screened, plus 59 for the seeds.

Stopping rule (stated before the search, applied after)

  • Effort bound: about 80 results screened in all.
  • Saturation: stop a family after two consecutive strings add no new included source.
  • Applied:
    • E: E16 added none, E17 added one. Not saturated.
    • F: F8 added none, F7 added one. Not saturated.
    • G: G33 added none, G34 added one. Not saturated.
  • The effort bound was exceeded (209 against 80), because each string returns about ten result blocks and 21 strings were needed to touch six sub-questions. No family stopped on saturation. Each stopped when every sub-question Q9a to Q9f had at least one source or a recorded "none found".
  • Consequence: Q9 is a first pass. A second pass on consulting firms' own documents and on accelerator guides would likely add sources.

Selection

  • Include: a source that states a rule or gives evidence on Q9a to Q9f, in English, about a pitch, a sales or executive presentation, a deck or a memo as its unit.
  • Exclude: entrepreneurship studies of crowdfunding pages or business plans read without a presentation; template vendors and AI slide-tool blogs repeating one another; pages with no producer.
  • Quality: grey sources scored on the 20 criteria of section 8 of protocol.md; a source under 10 serves as an example of practice only.

Limits

  • One reviewer, no second rater for these scores.
  • One personalising search engine; results of 2026-10-29 may not repeat.
  • Screening used titles and the engine's summaries. Numbers read only in a summary are marked "(secondary)".
  • Three of the formal sources (Chen 2009, Pollack 2012, Huang 2015) are paywalled; only abstracts were read. Brooks 2014's PNAS page sits behind a bot check that was not bypassed; the author copy in MIT DSpace was read in full.
  • Most investor studies are of US or UK angels, venture capitalists and televised pitches. Executive and board briefings have no study in this pool.

Q10 casual talks (added 2026-10-29)

Numbering continues the strands E (E19 on), F (F9 on) and G (G38 on). ERIC strings are numbered C1 to C10.

What a casual talk is, for this search: see the review, section 10.

Table 4 for Q10 (Garousi seven questions)

# Question Answer for casual talks
1 Complex subject, not solvable by the formal literature alone yes: meetup, lightning and community-conference norms are practice
2 Lack of volume, quality or consensus of evidence yes: almost no study measures what a lay listener keeps from a live talk
3 Context matters yes: voluntary attendance, lay or mixed audience, short slots
4 Goal to check research against practical experience yes
5 Goal to challenge practice rules with research yes: "tell a story", "use humour", "no slides", "avoid live demos"
6 Synthesis useful to both communities yes
7 Many practitioner sources yes: organiser handbooks, speaker guides, competition rules

Seven of seven are yes, so grey literature is included.

Q10 questions

  • Q10a Audience: what changes when attendance is voluntary and the audience is lay or mixed (attention, prior knowledge, motivation)?
  • Q10b Form: lightning talks and short fixed formats (PechaKucha, Ignite, five-minute talks): what is measured, what is practice?
  • Q10c Engagement: humour, storytelling, questions to the audience, a personal tone: what is measured, in which settings?
  • Q10d Slides or none: talks with few or no slides, or with a live demo.
  • Q10e Practice: meetup and conference-organiser guides, speaker guides for community talks, each scored.
  • Q10f Against: where casual-talk practice contradicts the evidence held in Q1 to Q9.

Engine and date

  • E, F and G strings: a general web search tool (one engine, US-based, returns a summary and links), run on 2026-10-29. A "screened" count is the number of result blocks returned. Each block was read as a title plus the engine's summary.
  • ERIC (https://api.ies.ed.gov/eric/, field numFound): phrase and Boolean search, education only, run on 2026-10-29. Titles of the hits were screened.
  • Metadata checks: the public Crossref REST API, called with no identity and no mailto.
  • OpenAlex counts: attempted on 2026-10-29 and answered HTTP 429 (daily budget used, retry after about 13 hours). No OpenAlex count exists for Q10.
  • Pages opened one at a time. No sign-in, no CAPTCHA, no bypass was tried. Two publisher pages (Springer) redirected to an authorisation step and were not followed.

E: existing reviews (Guideline 2)

Id String, as run Screened Included
E19 systematic review science communication public talks audience effects meta-analysis 9 1 (McCarthy 2025); a review of written science communication seen, not used (text, not talks)
E20 meta-analysis humor in instruction learning effect 10 0 (older meta-analyses named in a special-issue summary)
E21 Banas 2011 review of humor in educational settings four decades of research Communication Education 10 2 (Banas 2011, later read in full; Zhou 2025)
E22 PechaKucha systematic review education students presentations 9 0 (no review exists in these results)
E23 PechaKucha scoping review OR "systematic review" OR meta-analysis presentation format learning 9 0 (same result: primary studies only)
E24 audience response systems clickers meta-analysis learning outcomes lecture questions to audience 10 1 (Hunsu 2016, later read in full)
E25 seductive details meta-analysis learning Rey 2012 Educational Research Review 9 2 (Rey 2012, later read in full; Cheng 2026b, abstract read); Sundararajan 2020 already held
E26 personalization principle meta-analysis conversational style multimedia learning Ginns 2013 Mayer personalization effect 20 1 (Ginns 2013, later read in full); Mayer 2014c already held

F: formal

Id String, as run Screened Included
F9 what predicts TED talk views ratings analysis of TED talks study 9 1 (Segal 2026, a preprint)
F10 Sugimoto Thelwall TED talk presenters scientists popularizing science characteristics impact PLoS ONE 9 1 (Sugimoto 2013, full text)
F11 TED talks rhetorical analysis experiment persuasion audience gestures humor storytelling study views peer-reviewed 20 0 (no experiment; the gesture counts come from a popular source)
F12 storytelling narrative science communication effect on understanding engagement experiment Dahlstrom 10 1 (Dahlstrom 2014, full text)
F13 Fischer Thies science story good narrative transportation topic interest experiment 9 0 (abstract only through the engine; the full file could not be read)
F14 experiment humor in persuasive speech or lecture recall comprehension randomized relevant humor Garner humor in pedagogy 18 3 (Garner 2006, later read in full; Kaplan 1977 and Suzuki 2014, secondary)
F15 humour in science communication public audiences study effects on perceived scientist credibility engagement Riesch OR Yeo 18 2 (Frank 2025 read on the journal page; Freiling 2024, later read in full)
F16 science café OR "public lecture" audience survey who attends motivations evaluation study public engagement events 10 1 (Childers 2021, later read in full)
F17 talk without slides versus PowerPoint audience experiment recall engagement chalk talk blackboard study 9 0 (no study of a slide-free live talk)
F18 jargon barrier science communication lay audience experiment comprehension public talk Sharon Baram-Tsabari 20 1 (Rakedzon 2017, full text)
F19 vulgarisation scientifique conférence grand public revue de littérature effets sur le public 9 0 (French: course pages and essays; no evaluation study)

C: ERIC

Id String, as run Count Screened Included
C1 "PechaKucha" OR "Pecha Kucha" 21 21 0 new (Liao 2020 and Warmuth 2021 already held; no review)
C2 "lightning talk" OR "lightning talks" 5 5 0 (teaching reports)
C3 "Ignite talk" OR "Ignite talks" 1 1 0
C4 "TED talk" OR "TED talks" 85 30 0 (language-learning uses of TED talks as material; one multimodal analysis)
C5 "public lecture" AND science 5 0 0 (count only)
C6 "instructional humor" OR "humor in teaching" 26 0 0 (count only; reviews reached through E20 and E21)
C7 "science communication" AND ("public talk" OR "public lecture" OR "public engagement") AND ("systematic review" OR "meta-analysis") 0 0 0
C8 "five-minute" AND (presentation OR talk) AND (students) 13 0 0 (count only)
C9 "live demonstration" AND "lecture" 6 0 0 (count only)
C10 "vulgarisation scientifique" 0 0 0 (ERIC is English-language: zero says the engine is wrong for French, not that nothing exists)

G: grey

Types and producers (Guideline 6):

Type Producers Why
Organiser handbooks a professional society, a company community programme, a meetup blogger what organisers tell speakers
Speaker guides community conference advice pages, lightning-talk tips what speakers tell each other
Competition rules Three Minute Thesis and its French form Ma thèse en 180 secondes the best-known fixed short format for lay audiences
Format owners PechaKucha, Ignite, TED already held in Q6, linked
Id String, as run Screened Opened Included
G38 Three Minute Thesis 3MT study effects on doctoral students communication skills audience evaluation 19 0 0 (programme pages; no outcome study)
G39 Ma thèse en 180 secondes règlement concours conseils jury critères 10 2 0 (the official PDFs opened returned HTTP 404; rules known only from the engine's summary)
G40 how to give a lightning talk guide conference organisers speaker advice five minutes 10 0 0 (blog tips seen in a summary, not opened)
G41 PyCon speaker guide talk proposal advice community conference speaker handbook audience beginners 18 1 1 (PyCon 2026)
G42 meetup organizer guide how to find speakers format talk length Q&A community tech meetup handbook 9 3 3 (ACM Local 2026, WooCommerce 2026, Hilton 2022)
G43 Three Minute Thesis OR "lightning talk" OR "short presentation" time limit audience comprehension experiment short versus long talk retention 18 0 0 (no controlled comparison of short against long talks)

Totals: E 86, F 141, G 84 = 311 result blocks screened, plus 57 ERIC titles.

Stopping rule (stated before the search, applied after)

  • Effort bound: about 80 results screened in all.
  • Saturation: stop a family after two consecutive strings add no new included source.
  • Applied:
    • E: E22 and E23 added none, and the next string (E24) added one; the PechaKucha strand saturated on "no review exists".
    • F: F16 added one, F17 and F19 added none; F18 added one. Not saturated.
    • G: G38 to G40 added none in a row, then G41 and G42 added four. The practice strand did not saturate.
  • The effort bound was exceeded (311 against 80), as in Q9, because each string returns about ten blocks and 25 strings were needed to touch six sub-questions. A family stopped when each of Q10a to Q10f had at least one source or a recorded "none found".
  • Consequence: Q10 is a first pass. Organiser guides of public-engagement bodies and the primary studies of Q10c (humour, storytelling) behind the secondary reads would likely add sources.

Selection

  • Include: a source in English that states a rule or gives evidence on Q10a to Q10f. It is about a talk, a lecture or a short presentation to a voluntary or lay audience, or about the device itself: humour, narrative, personal tone, audience questions.
  • Exclude: language-learning uses of TED talks as listening material; marketing surveys of content formats; pages with no producer.
  • Quality: grey sources scored on the 20 criteria of section 8 of protocol.md; under 10 of 20 serves as an example of practice only.

Limits

  • One reviewer, no second rater for these scores.
  • One personalising search engine; results of 2026-10-29 may not repeat.
  • Screening used titles and the engine's summaries. Numbers read only in a summary are marked "(secondary)".
  • Seven of the 17 new formal sources were secondary. Five were later read in full (Banas 2011, Garner 2006, Freiling 2024, Hunsu 2016, Rey 2012); Kaplan 1977 and Suzuki 2014 remain secondary. An anonymous request to the Springer pages of Cheng 2026b and Childers 2021 returned their abstracts.
  • The studies of humour, narrative and personal tone are of text, recorded lectures, lessons on computer, screenshots of posts or surveys, not of live talks to voluntary audiences.
  • No query returned a study that measures what a lay audience keeps from a live casual talk.