Five sources agreeing on a fact looks like strong evidence. It usually isn’t.
When we’re researching a topic and the same statement keeps surfacing, the natural reaction is to feel more confident. Each new appearance reads like a vote. By the time we’ve seen it in five places, we treat it as established. That instinct is one we’ve had to learn to mistrust, because the relationship between “number of sources” and “amount of evidence” is much weaker than it feels.
The shape of borrowed agreement
Most reporting and most writing online is downstream of something else. A primary report comes out, ten outlets summarize it, twenty more summarize the summaries, and the next week a hundred articles repeat the same paragraph almost verbatim. From the surface this looks like wide coverage. Underneath, it’s one source spreading.
We notice this most clearly with statistics. A number gets cited. The article we’re reading attributes it to a previous article. That article attributes it to a report. The report cites a study. The study has a footnote pointing to another paper. Eight of those layers can sit between us and whoever first put the number on a page, and most of them won’t have done the work to verify it. They trusted the layer above.
When we’re working from a stack of sources that all say the same thing, the question that matters isn’t how many sources. It’s how many independent origins.
Tracing back
Our default move when something looks too tidy is to walk the citation chain backward. We pick the strongest claim in the article, the one our finding would lean on hardest, and find the source it points to. We open that source and find the source it points to. We keep going until we hit a wall: an unattributed statement, a primary report, a researcher’s blog post, a press release, a court filing, a benchmark.
Once we hit the bottom, we ask whether the wall is load-bearing. A government statistical release usually is. A press release from a company describing its own performance usually isn’t, no matter how many trade publications repeated it.
What we find at the bottom often surprises us. Five articles can collapse into one report whose authors never claimed what’s now being repeated about them. A statistic can turn out to come from a single chart in an old presentation that nobody else has actually checked. A “consensus view” can dissolve into one influential blog post quoted respectfully by everyone who came after.
What independence actually buys
When two sources arrive at the same conclusion through visibly different methods, looking at different data, with different incentives, that’s information. Two independent origins reaching the same place is real evidence. We weight that pair higher than ten downstream restatements.
Independence is hard to verify but easy to falsify. If two sources cite each other, they aren’t independent. If they share authors or institutional ties, we discount the agreement. If one is summarizing the other, we collapse them into one. By the time we’re done, what looked like a chorus of ten can shrink to two or three voices, sometimes one.
We’ve learned to write our findings with that in mind. Instead of “multiple sources confirm,” we say “the [name] report, confirmed independently by [other].” When we can only find one origin, we say so plainly: “This number originates with one report by X. Other articles repeat it without independent verification.” That sentence sounds weaker than the alternative, and it is, because the underlying evidence is weaker.
When everyone agrees and we still can’t trace it
The trickier case is the assertion that everyone seems to know but nobody clearly attributes. The kind of thing that shows up in articles as “it’s well known that…” or “experts agree that…” with no link. These statements have the texture of consensus, and sometimes they reflect one. Often they reflect a folk belief that started somewhere and lost its origin along the way.
When we find one of these, we don’t strip it from the findings. We flag it. The note we leave for whoever reads our work is something like: “This is widely repeated. We could not find a primary source. Treat as conventional wisdom, not established fact.” That gives the reader the latitude to make their own call. If the claim is load-bearing for whatever they’re building on top of our research, they know to dig further. If it’s atmospheric, they know not to over-rely on it.
What changes about confidence
The thing we’ve internalized, slowly, is that confidence in a finding should track source independence and not source count. Five repetitions of the same source is the same evidence as one. Two truly independent origins reaching the same conclusion is more than five repetitions of either.
This adjustment changes how the work feels. The first time we find what looks like overwhelming agreement and trace it back to a single thin reed, the instinct is to feel cheated. Eventually it becomes routine. The ease with which information propagates online is part of the environment we research in, and the research has to account for it. The world doesn’t owe us a clean correlation between popularity and truth.
What changes after a while is the moment when we stop counting sources. We start asking, instead, what the sources are tied to and whether those ties are load-bearing. The number of voices in the chorus matters less than where the song came from.