Follow the Citation Chain · Method

Purpose

Most numbers and quotes reach you secondhand: a post cites an article, the article cites a report, the report cites a study. Following the citation chain means walking those links back, one at a time, until you reach the source that actually produced the claim. Then you compare the end of the chain with the start.

The original almost always says something narrower than the version you heard. Each retelling drops a condition, rounds a number, or widens the claim. You only see it by going back.

This is the last of the four moves for checking a claim. In a research project it isn’t a separate step at all: you do it while you read, every time a reference carries weight in your argument. Umberto Eco’s advice is the same. When something you’re reading cites a source that matters, go back to that source and check it before you rely on it.

When you need it

A statistic arrives with “studies show” or “research says” and no link.
A source cites another source: a post quoting an article, which quotes a report.
A number sounds too round, too strong or too convenient.
A quote is credited to a famous person, especially one who died before the internet.
You’re about to repeat the claim yourself.

How to do it

1Write down the exact claim

Copy the words, the number, who it’s credited to, and where you saw it. Vague claims can’t be traced, and you’ll need the exact wording later to see what changed.

2Find what it cites

Look for a link, a footnote, or “according to.” If there’s nothing, search for the exact number or phrase in quotation marks with one distinctive word from the claim. The earliest dated result is often the next link back.

3Go back one link, then repeat

At each link, note the source, its date, what it actually says, and what it cites in turn. Keep going until you reach a source that cites nothing further because it produced the information: the study, the dataset, the transcript, the bill, the speech.

4Get the original itself

A DOI (a paper’s permanent ID, like 10.1037/h0024532) opens the publisher’s page at doi.org/ plus the ID. If it’s paywalled, check Google Scholar for a free version, ask your library, or email the author. Authors are usually glad to send a copy.

5Read where the claim lives

Find the passage the chain depends on, and read the methods around it: who was studied, what was measured, when, and under what conditions. The limits of a finding live in the methods, rarely in the abstract.

6Compare the end with the start

Lay the original next to the claim you started with. What changed: the scope, the number, the certainty, the wording, or who said it? Then cite the original, and if you couldn’t reach it, say so (“cited in…”) instead of citing a source you haven’t read.

Try it

Your task

“93% of communication is nonverbal.” The line turns up in sales trainings, public-speaking courses and management books, usually credited to research by a psychologist. Follow the chain back. Find the original research, and find out what it actually measured.

Show a hint

The name attached to the claim is Albert Mehrabian. Check whether he has his own website, and look for papers he published in 1967.

Show the answer and the path

The chain runs from popular retellings, to Mehrabian’s book Silent Messages (his website points to pages 75 to 80), to two papers from 1967:

Mehrabian and Wiener, “Decoding of Inconsistent Communications,” Journal of Personality and Social Psychology 6(1): 109–114.
Mehrabian and Ferris, “Inference of Attitudes from Nonverbal Communication in Two Channels,” Journal of Consulting Psychology 31(3): 248–252.

What the studies measured was narrow: how listeners judged a speaker’s feelings (liking or disliking) from single spoken words, where the tone of voice and the facial expression sometimes contradicted the word. One study used the word “maybe.” The famous split (7% words, 38% tone, 55% face) came from combining the two.

What changed along the chain: a finding about judging feelings from mismatched one-word messages became a rule about all communication. Mehrabian himself says on his website that the equations don’t apply unless the speaker is talking about their feelings or attitudes. The idea that tone and expression matter survives; the 93% doesn’t.

Common mistakes

Stopping at the first source that sounds official. A news story, a think tank or a textbook citing someone else is still a middle link, not the end.
Trusting a citation without reading it. In medical journals, about one in four citations doesn’t support the statement it’s attached to (Jergas and Baethge, 2015). Citations get copied from other papers’ reference lists without anyone opening the source.
Reading only the abstract or the press release. The conditions that limit a finding are in the methods section.
Mixing up versions. Reports get revised and datasets get updated. Note the date and edition at each link.
Treating a dead end as a debunking. “I couldn’t find the source” means the claim is unsupported so far. It doesn’t mean the claim is false.

Limits

Some chains can’t be finished. Pages disappear (try the Wayback Machine), sources sit behind paywalls, and some claims rest on unpublished data or “personal communication.” Say where your chain stops instead of guessing past it.

Reaching the original tells you what it says, not whether it’s right. Original studies can be small, badly designed or wrong, and a claim with a broken citation can still turn out true. Judging the evidence is a different job: ThinkingCritically.net covers it.

For a passing claim, going two links back is often enough to see whether the version you heard has drifted. Save the full trace for claims you plan to repeat, rely on or publish.