Reading persuasion research: what replicates and what…

Transcript of the narration, chapter by chapter.

0:00 Reading persuasion research

Stand in a power pose for two minutes before a big pitch and your hormones will change and you'll close more deals. You may have heard that. It came from a famous study and a very popular talk. And its hormonal effects failed to replicate, and one of the original authors later said publicly she no longer believes the effect is real. In this lecture you'll learn how to read persuasion research like a professional: what the replication crisis was, which persuasion effects are robust and which aren't, five questions to ask about any science-says claim, and how to test in your own context.

0:45 Why it matters

Why does a seller need this? Because sales training, marketing blogs and pitch decks are full of science-says claims. People buy more with fewer options. Willpower runs out after too many decisions. A particular word lifts compliance by a precise percentage. Some of these are well supported. Some come from single small studies that later failed. If you repeat them, you can mislead clients, your team and yourself. And in a sales context, quoting an invented or unsupported statistic can itself be a misleading claim. Research literacy is part of honest selling.

1:25 The replication crisis

Here's what happened. From around twenty eleven, psychologists began systematically re-running well-known experiments. In twenty fifteen, a large collaboration tried to replicate one hundred published psychology studies, and fewer than half produced a statistically significant result again, with effects on average around half the original size. Then Many Labs projects ran the same experiments across dozens of labs. Some effects, like anchoring, replicated strongly. Others didn't. The causes are well understood: small samples, flexible analysis, and journals favouring striking results. The field has responded with pre-registration, bigger samples and multi-lab studies. Think of it like a restaurant review site that discovers many five-star reviews came from the owners' friends. The good restaurants are still good. You just need better ways to tell which ones they are.

2:20 Evidence map (simplified)

So what's robust for persuasion, and what isn't? A simplified map. Robust: anchoring, default effects and framing. Generally effective but wording-sensitive: social norms messaging. Broadly replicated but with debated size: loss aversion. Conditional: choice overload, the famous jam study, where a twenty ten meta-analysis found an average effect near zero. Weaker in realistic settings: the decoy effect. And largely failed: ego depletion, the hormonal effects of power posing, and many social priming findings, where subtle cues supposedly changed behaviour. Even nudges as a whole are debated: a twenty twenty-two meta-analysis found a positive average effect, but a re-analysis adjusting for publication bias found little evidence of an average effect, while defaults look more robust.

3:10 Five questions

Here are five questions to ask about any science-says claim. One: what's the original source? A named study, author and year, or a blog quoting a blog? Two: how big was the sample, and who was in it? Forty psychology students at one university isn't people. Three: has it been replicated, ideally pre-registered, by independent teams? Four: how big is the effect in real-world conditions? Lab effects often shrink in the field. And five: does it apply to my buyers, my market and my decision type? A grocery-shelf study may not predict a business software purchase.

3:52 Reading research: key terms

A few terms will help you read research quickly. A meta-analysis combines many studies to estimate an average effect, which is usually more reliable than any single study, although it can be distorted if unpublished null results are missing. That's called publication bias. Pre-registration means researchers publicly record their hypothesis and analysis plan before collecting data, which makes it much harder to go fishing for a striking result. A registered replication report is a pre-registered attempt, often across many labs, to repeat a famous finding. And effect size tells you how big a difference is, not just whether there is one. A statistically significant effect can still be tiny in practice. When a claim rests on a pre-registered, multi-lab result with a meaningful effect size, take it seriously. When it rests on one small study, be cautious.

4:52 Example 1: deck slide

First example, the simple one: a sales-deck slide. Before: science proves that offering three options increases conversion by thirty per cent. Let's ask the questions. Source? None given. Replication? The famous choice-overload research is mixed. Relevance? Unknown. So here's the after version. Research on choice overload is mixed: simplifying options seems to help most when choices are complex or unfamiliar. For your pricing page, we suggest testing a three-tier layout against your current one. It's less exciting. It's honest, it's defensible, and it leads straight to a test.

5:30 Example 2: Dubai agency (illustrative)

Now a realistic business scenario, with illustrative details. A Dubai agency's pitch deck promised clients a guaranteed twenty per cent uplift from scarcity messaging, backed by science. A client's analyst asked for the source. There wasn't one. It had been copied from a blog, which cited another blog. Embarrassing, and potentially misleading. So the agency rebuilt its method. It cites only well-replicated principles, carefully. It labels illustrative numbers as illustrative. And it proposes A B tests with success measures agreed in advance. Illustratively, it lost one prospect who wanted guarantees, and won two who valued the honesty. And its case studies are now based on its own tested results.

6:17 Watch me do it: evidence check

Watch me do it. I'll run the evidence-check template on a claim I hear constantly: buyers make worse decisions late in the day because of decision fatigue, so book your closing calls in the afternoon. Original source? It traces back to ego-depletion research, the idea that willpower runs out like fuel. Sample and setting: mostly lab studies with students. Replications: two large pre-registered multi-lab replications found little or no effect. Effect size in the field: unclear. Relevance to my context: low. Decision: drop it. And if I'm curious about call timing, I'll test it in my own pipeline instead: alternate morning and afternoon slots for two months, track show-up and next-step rates, and report the result honestly, even if it's no difference.

7:10 Test in your own context

That leads to the most reliable evidence of all for your business: a well-designed test in your own audience. Change one thing at a time. Make sure you have enough volume to tell a real difference from noise. Agree the metric before you start, so you don't pick whichever number looks best afterwards. And report honestly, including when the answer is no difference. That's the same discipline that fixed psychology's replication problems, applied to your pricing page, your emails or your call scripts.

7:46 Mistakes + measures

Common mistakes. Quoting statistics without a source. Citing a single small study as settled science. Assuming lab effects transfer to real buyers unchanged. Promising clients specific uplifts because science says so. And cherry-picking your own test results. How to measure your own research hygiene: the share of claims in your decks and content that have a traceable source, the share labelled illustrative when they are, and the number of claims you've dropped after checking. Dropping a claim isn't failure. It's quality control.

8:22 Recap + try this now

Recap. Some famous persuasion findings are robust, like anchoring, defaults and framing. Others, like ego depletion, power posing and social priming, largely failed to replicate. Ask five questions: source, sample, replication, real-world size and relevance. Never quote invented or unsupported statistics, and test in your own context with honest reporting. Your try-this-now action: pick one science-says claim you've heard or used, run it through the evidence-check template, and decide: use, caveat, test or drop. Next, we'll look at Cialdini's principles of influence, with this lens switched on.

Why sellers need research literacy

Sales training, marketing blogs and pitch decks are full of "science says" claims: people buy more with fewer options, a power pose makes you more persuasive, willpower runs out after too many decisions, a specific word lifts compliance by a precise percentage. Some of these are well supported. Some come from single small studies that later failed to replicate. Repeating them can mislead your clients, your team and yourself — and quoting an invented or unsupported statistic in a sales context can itself be a misleading claim.

The replication crisis in brief

From around 2011, psychologists began systematically re-running well-known experiments. In 2015 the Open Science Collaboration attempted to replicate 100 published psychology studies and found that fewer than half produced a statistically significant result again, with average effect sizes roughly half those originally reported. Large "Many Labs" projects ran the same experiments across dozens of labs: some effects (such as anchoring) replicated strongly; others did not. The causes are well understood — small samples, flexible analysis, publication bias towards striking results — and the field has responded with pre-registration, larger samples and multi-lab studies.

A quick evidence map for persuasion topics

Topic	Evidence picture (simplified, 2026)
Anchoring	Robust across many replications
Default effects	Consistently strong across domains
Framing (gain/loss wording)	Widely replicated
Social norms messaging	Generally effective, but exact wording effects vary (e.g., mixed replication of the "most guests reuse towels" boost)
Loss aversion	Core finding of prospect theory, broadly replicated across countries; its size and generality are debated
Choice overload ("jam study")	Average effect near zero in a 2010 meta-analysis; appears under specific conditions
Decoy effect	Weaker or absent with realistic choices in later work
Ego depletion	Large pre-registered replications found little or no effect
Power posing (hormonal effects)	Failed to replicate; one original author publicly said she no longer believes the effect is real
Social priming (subtle cues changing behaviour)	Many high-profile findings failed to replicate
"Nudges" overall	A 2022 meta-analysis reported a positive average effect; a re-analysis adjusting for publication bias found little evidence of an average effect, though some interventions (such as defaults) look more robust

Treat this as a starting point; check recent reviews before relying on any single effect.

Five questions to ask about any "science says" claim

What's the original source? A named study, author and year — or a blog quoting a blog?
How big was the sample, and who was in it? Forty psychology students in one university is not "people".
Has it been replicated — ideally pre-registered, by independent teams?
How big is the effect in real-world conditions? Lab effects often shrink in the field.
Does it apply to my buyers, market and decision type? A grocery-shelf study may not predict a B2B software purchase.

Hands-on: an evidence check for a claim you want to use

CLAIM I want to use: "______"
Original source (author, year, journal): ______
Sample and setting: ______
Replications / meta-analyses found: ______
Effect size in field settings (if known): ______
Relevance to my context: high / medium / low
DECISION: use as stated / use with caveat / test ourselves / drop
Safe wording: "Research suggests… in some contexts" / "In our own tests…"

Before and after: a sales-deck slide

Before: "Science proves that offering 3 options increases conversion by 30%!"

After: "Research on choice overload is mixed: simplifying options seems to help most when choices are complex or unfamiliar. For your pricing page, we suggest testing a three-tier layout against your current one."

Worked example

A Dubai agency's pitch deck promised clients a "guaranteed 20% uplift from scarcity messaging, backed by science". A client's analyst asked for the source; there wasn't one. The agency rebuilt its method: it cites well-replicated principles carefully, labels illustrative numbers as illustrative, and proposes A/B tests with pre-agreed success measures. It loses one prospect who wanted guarantees — and wins two who valued the honesty.

Test in your own context

The most reliable evidence for your business is a well-designed test in your own audience: one change at a time, enough volume, a pre-agreed metric, and honest reporting — including when the result is "no difference".

Reading persuasion research: what replicates and what doesn't