Article: Why Some Messages Convert
Reviewed: November 2026 · Method version 2 · How these reviews work
Each factual claim in the article was checked against the source it cites, then challenged by reviewers whose job was to find the strongest evidence against it: replication attempts, meta-analyses, later papers that cite the original work, and research from other fields that study the same mechanism. A separate reviewer weighed both sides, and a script re-checked every quoted source. Where a claim was overstated or misattributed, the article was corrected. The research below supports the article as it now reads.
When people are motivated and able to think about a message they weigh its arguments and weak arguments lose; when they aren’t, they lean on simpler cues such as who is speaking. Petty and Cacioppo called these the central and peripheral routes to persuasion.
The two-route model has broad meta-analytic support (Carpenter 2015). A failed Many Labs 3 replication concerned a different ELM finding (need for cognition), and a later nine-site replication recovered it at a smaller size.
In the 1983 test, Petty, Cacioppo and Schumann showed undergraduates a magazine ad for a product under high or low involvement.
Design details match the 1983 abstract.
The ad carried either strong or weak arguments and was endorsed either by well-known sports celebrities or by ordinary people.
Design details match the 1983 abstract.
Argument quality changed attitudes more when involvement was high.
Meta-analytic support (Carpenter 2015); replication effects are smaller than in the original studies.
The famous endorsers mattered more when involvement was low.
Replicated in the 1981 companion study; endorser effects in general depend on moderators such as fit (Knoll and Matthes 2017).
Shu and Carlson found that when people knew the source wanted to persuade them, impressions of a product improved with each positive claim up to three.
Replicated by a 2021 Journal of Consumer Psychology study under the no-imagery condition.
At the fourth, their knowledge of persuasion kicked in and they grew skeptical of all the claims, including the good ones. A later study found a limit: when readers were picturing themselves using the product, more than three claims helped.
A 2021 replication found the charm of three only without mental imagery; with imagery, more claims raised evaluations. Weaver, Garcia and Schwarz (2012) offer a dilution account of similar drops.
Sources with no persuasion motive showed no such peak.
Reported in the first of Shu and Carlson’s four experiments.
The authors call it the “charm of three.”
Term confirmed in the abstract.
Quality persuades the reader who is paying attention, and that same reader, who knows they are being sold to, reads volume as salesmanship. The skimmer is a different case: in Shu and Carlson’s own test, readers under cognitive load stopped penalizing a fourth claim.
Shu and Carlson (2014) found the drop after a third claim only for sources with a persuasion motive, and their cognitive-load experiment (reported here from a 2015 Frontiers in Psychology description and a 2020 Psychology Today summary; the full text was not accessible) removed the penalty for a fourth claim. A 2021 Journal of Consumer Psychology replication found more than three claims helped readers engaged in mental imagery, a 2015 Frontiers in Psychology study found more positive claims persuaded more even as persuasion knowledge rose, and a 2025 Personality and Social Psychology Bulletin paper found many arguments raise perceived persuasive intent and perceived expertise at once, with little net effect. Campbell and Kirmani (2000) show that even busy readers use persuasion knowledge when the sales motive is obvious, which supports the article’s premise for attentive readers. Under low involvement, the number of arguments can itself act as a favorable cue (Petty, Cacioppo and Goldman 1981). A 2022 Social Science & Medicine experiment (n = 4,616) found three stated benefits beat one for a public-health policy, a source without an obvious sales motive, which fits Shu and Carlson’s boundary rather than contradicting it.
Packard and Berger ran five studies, including a text analysis of more than 1,000 real conversations between customers and employees.
Study scope confirmed in the abstract.
Customers were more satisfied, more willing to buy and bought more when employees spoke concretely.
Later advertising experiments find the concreteness advantage is largest under uncertainty (International Journal of Advertising 2023).
This happened because customers took concrete words as a sign that the employee was listening.
Mechanism confirmed in the abstract; concreteness also raises judged truth (Hansen and Wänke 2010).
On a page, specifics likely do a similar job, especially for readers who are unsure and comparing options.
Packard and Berger (2021) studied one-to-one service talk. For ads, concrete wording helps most under uncertainty (International Journal of Advertising 2023).
Past three claims, extra bullets start to cost credibility across the whole list, and weaker ones dilute the strong ones.
Supported in direction by Shu and Carlson (2014) and the dilution findings of Weaver, Garcia and Schwarz (2012); the per-bullet framing and the transfer to product feature lists are untested.
At the desk, vague popularity cues count for less and checkable facts count for more. A cue the reader can verify, such as a named award or a review count, still carries weight. This is the section where argument quality matters most.
Petty, Cacioppo and Schumann (1983) found the endorser cue mattered less, not nothing, under high involvement, and a meta-analysis of 134 studies (Carpenter 2015) confirms argument quality matters more under central processing; a 2026 replication in advertising found the same for human-written ads but did not find the endorser-by-involvement interaction. Chaiken and Maheswaran (1994) found a credibility cue still shaped attitudes under high importance when arguments were ambiguous, Pierro and colleagues (2004) traced the reduced cue effect to cues seeming less relevant, Darke and colleagues (1998) found highly motivated people used consensus information when the poll was reliably large, and Park, Lee and Han (2007) found high-involvement shoppers still responded to review counts when review quality was high. Review volume raises sales (You, Vadakkepatt and Joshi 2015), a popularity ranking raised restaurant demand 13 to 20 percent in a field experiment (Cai, Chen and Fang 2009), and awards and quality seals outperform celebrity endorsers (Knoll and Matthes 2017).
Magic Words (Jonah Berger, Harper Business, 2023, 256 pages) starts from the premise that certain words are more impactful than others; Berger co-authored the concreteness study.
Book details confirmed through the author page and Open Library.
These edits were made on November 12, 2026 after the review.
Claim reviewed: Ads, subject lines and homepages are mostly met on the peripheral route; pricing pages, comparison tables and checkout are where the central route takes over, because money or effort is about to be spent.
Now reads: Online, the split shows up often. Ads, subject lines and homepages are usually met by people doing something else, so they tend to get the peripheral route. Pricing pages, comparison tables and checkout are more likely to get the central route, because money or effort is about to be spent. The route belongs to the reader, not the page, as the train and the desk show.
Why: No source is given, and no study tests whether whole page types get read one way or the other. How carefully someone reads depends on their own state, which the article’s train-and-desk story shows: the same homepage got both kinds of reading. The mapping is a reasonable rule of thumb but should be framed as one.
Claim reviewed: Copy that converts usually knows which of the two routes it is written for.
Now reads: Copy that converts starts by deciding which of the two it is written for.
Why: This is the article’s thesis, stated as a fact about conversion, but no source is cited and pass 3 found no conversion data testing it. Lab work on the two routes is consistent with the idea, but that is a step away from conversion rates.
Claim reviewed: At the fourth claim, people’s persuasion knowledge kicked in and they grew skeptical of all the claims, including the good ones.
Now reads: At the fourth, their knowledge of persuasion kicked in and they grew skeptical of all the claims, including the good ones. A later study found a limit: when readers were picturing themselves using the product, more than three claims helped.
Why: The source says exactly this, including skepticism toward all the claims. But a 2021 replication found the effect reverses when readers imagine using the product, where more than three claims helped. A separate line of research explains similar drops by weak items diluting strong ones rather than by skepticism. Now cites When the ‘Charm of Three’ Fades: Mental Imagery Moderates the Impact of the Number of Ad Claims on Persuasion, Journal of Consumer Psychology, 2021.
Claim reviewed: Volume reads as salesmanship to anyone who knows they are being sold to, which on a marketing page is everyone.
Now reads: Quality persuades the reader who is paying attention, and that same reader, who knows they are being sold to, reads volume as salesmanship. The skimmer is a different case: in Shu and Carlson’s own test, readers under cognitive load stopped penalizing a fourth claim.
Why: Three separate labs now find that piling on claims makes readers sense a sales pitch, so the core of the sentence holds for readers who are paying attention and know they are being sold to. But ‘anyone’ and ‘everyone’ reach too far: Shu and Carlson’s own cognitive-load test removed the fourth-claim penalty for distracted readers, a 2021 replication found extra claims helped readers who pictured the product, and a 2025 study found many arguments also make the source look more expert, leaving little net loss. The sentence is right about the attentive reader and wrong to include the skimmer the article opens with. Now cites Fifteen Reasons You Should Read This Paper: How Providing Many Arguments Increases Perceptions of Both Expertise and Persuasive Intent, Personality and Social Psychology Bulletin, 2025.
Claim reviewed: Past three claims, each new bullet costs credibility across the whole list.
Now reads: Past three claims, extra bullets start to cost credibility across the whole list, and weaker ones dilute the strong ones.
Why: Shu and Carlson show a peak at three and lower ratings at four or more, and dilution research agrees that weaker items drag down strong ones. But ‘each new bullet’ implies a steady loss per item that the source does not report, the studies used short ad claim lists rather than feature lists, and the effect reversed for readers imagining the product. Now cites Weaver, Garcia & Schwarz, The Presenter’s Paradox, Journal of Consumer Research 39(3), 2012.
Claim reviewed: On a page, specifics do the same job: concrete copy shows you understood the reader’s situation, while abstract value words show nothing.
Now reads: On a page, specifics likely do a similar job, especially for readers who are unsure and comparing options.
Why: Packard and Berger studied employees replying to individual customers, so ‘the same job’ on a one-to-many web page is an analogy, not a tested result. Related work does favor concrete wording in ads, especially for uncertain readers, and concrete statements are judged more likely true. The point is sound, the mechanism on a page may differ. Now cites When it pays to be clear: the appeal of concrete communication under uncertainty, International Journal of Advertising, 2023.
Claim reviewed: At the desk (central route), popularity cues are worth little and checkable facts are worth a lot; this is where argument quality decides the outcome.
Now reads: At the desk, vague popularity cues count for less and checkable facts count for more. A cue the reader can verify, such as a named award or a review count, still carries weight. This is the section where argument quality matters most.
Why: The cited study and a meta-analysis of 134 studies agree that argument quality matters more when readers are engaged, and a 2026 replication found the same for human-written ads. But ‘worth little’ overstates it: under high involvement a credibility cue still moved attitudes when the arguments were ambiguous, motivated readers used popularity information when it was reliable, and high-involvement shoppers still responded to review counts when review quality was high. Vague lines like ‘Trusted by industry leaders’ probably are weak; a checkable award or review count is not. Now cites Darke, Chaiken, Bohner, Einwiller, Erb & Hazlewood, Accuracy Motivation, Consensus Information, and the Law of Large Numbers, Personality and Social Psychology Bulletin 24(11), 1998.