Trace a Claim to Its Primary Source

In the last chapter you learned to check a source against the wider world by reading laterally. But finding other pages that repeat a claim is not the same as knowing the claim is true. They may all be copying the same shaky original. The fourth SIFT move, Trace, fixes that: you follow a claim back upstream, from where you found it all the way to the first-hand evidence it rests on, and check whether that evidence really says what everyone claims. This chapter shows you how, and what to do when the trail goes cold.

Follow a Claim Back to Where It Started

Here is something you will see constantly: a claim arrives with no original attached. “Studies show,” “a report found,” “the city says.”

Most claims reach you secondhand. A post quotes an article, the article quotes a press release, the press release describes a study, and the study sits at the bottom of the pile where almost no one looks. Your job, and soon your habit, is to follow that claim back to where it started.

Tracing means walking that chain in reverse: from the version in front of you, back through each source that passed it along, to the primary source, the original study, dataset, or document where the claim first appeared.

Because claims drift as they travel. A careful “risk rose 2% in one small trial” becomes “policy slashes risk” three retellings later. The only way to know what was really found is to read the original, not the retelling.

Picture a headline: “New bike-lane policy cut accidents by 40%.” Where does the 40% come from? The article links to a news brief, the brief links to a council press release, and the release mentions a report. Each of those is a messenger, not the source, so you keep going until you reach the report itself and read the actual number.

Maybe the report says accidents fell 40% on three streets over two months. Suddenly the headline means something far narrower. The move is always the same: do not stop at whoever repeated the claim, find whoever originated it.

Trace a Claim, and Document a Broken Trail

You can trace almost any claim with the same short routine. Work down it slowly, and write down each step, because the record is what you will hand in for your project.

• Start with the claim exactly as you found it, and note where you found it.

• Find the link or reference it leans on, and follow it one step upstream.

• Repeat: at each new source, ask whether it originated the claim or just repeated it, and keep going until it stops pointing elsewhere.

• Read the primary source, the original study, dataset, or document, and check whether it actually says what the claim says.

• If the trail breaks, a dead link, no citation, an “original” that does not exist, record exactly where it ends and stop.

See the Provenance Chain

It helps to picture tracing as a chain. Each link is a source, and you walk from right to left, from the version you met back toward the origin.

When you reach the primary source, the claim is traceable. When a link is missing, a dead link, a citation that leads nowhere, the chain breaks, and that break point is your documented “untraceable” result.

Trace One Claim End to End

Let’s follow a real-looking civic claim all the way down. You read a post: “The city’s new recycling program cut landfill waste by 60%.” That is the kind of number worth checking before you repeat it.

Step 1, where you found it. A community Facebook post, no source, just the figure and a celebration emoji.

Step 2, one step upstream. The post links to a local news article. The article says the same 60% and credits “a city council announcement.”

Step 3, keep going. You find the council’s press release. It says landfill waste “from the pilot neighborhoods dropped 60% over the first quarter.” Narrower already, a pilot, one quarter, not the whole city.

Step 4, reach the primary source. The release links to the council’s data report. The report shows the 60% figure covers three blocks over three months, with a footnote that two blocks already had private composting. So the primary source exists, but it measures something far smaller than “the city cut waste 60%.”

Verdict: traceable, but exaggerated. The number is real; the headline stretched it. That composting footnote matters: some of the drop may not be the program at all, so even the pilot figure deserves caution. You would record the full chain and note that the claim is supported only for the pilot, not the city, and is partly confounded.

Your Turn!

Time to trace a claim yourself and record where it leads. Below is a claim and the chain you uncover when you start clicking upstream.

The claim. A blog states: “Remote work boosts productivity by 35%, according to a major study.”

What you find when you trace it:

• The blog links to a newsletter.

• The newsletter links to a company’s marketing page.

• The marketing page says “internal research shows up to 35%” and links to nothing.

• No published study with that figure appears anywhere you search.

Your deliverable. Write the chain as four short lines, claim → newsletter → marketing page → origin, then a one-sentence verdict: is the claim traceable to a primary source, or does the trail break, and where? When you are done, check your answer against the model solution at the end of this chapter.

Let’s Recap!

• Tracing means following a claim upstream from where you found it to the primary source, the original study, dataset, or document it rests on.

• Each link that only repeats a claim is a messenger, so keep going until you reach the origin or run out of links.

• Claims drift as they travel, so reading the original often reveals it is narrower than the headline suggested.

• An untraceable claim is a valid, documented result: record exactly where the trail breaks instead of treating it as failure.

• The chain you record, source by source to the origin or break point, is what your project’s Claim-Tracing deliverable asks for.

You can now follow a claim down to its evidence, or prove there is none. In the next chapter you will learn to spot the tricks that make weak evidence look strong: cherry-picked data, correlation dressed up as cause, and the untraceable “studies show.”

Model Solution

This is the solution to the “Your Turn!” activity above. Trace the claim yourself before reading on.

The chain. Claim (“remote work boosts productivity 35%”) → newsletter (repeats it) → company marketing page (“internal research shows up to 35%”) → origin: none, no published study, no link, nothing searchable.

Verdict. The claim is not traceable to a primary source; the trail breaks at the marketing page. The only “evidence” is an unnamed internal claim by a party with a reason to make remote work look good. You would record “untraceable, no primary source,” which means the 35% cannot be relied on, however confidently it is stated.