How to read a discovery announcement

You do not need to be an astronomer to tell a strong result from an oversold one. Most of the signal is on the surface, in the language, and it is the same handful of tells every time.

Here are six questions worth asking of any discovery announcement.

1. Has anyone outside the team checked it?

Look for whether it is a paper or a preprint [3]. A preprint is the authors saying "we think this"; a peer-reviewed paper is that claim having survived people whose job was to find the hole in it.

Neither guarantees truth. But a preprint reported as a discovery has been promoted a full level of confidence by someone other than its authors, and that someone is usually the person writing the headline.

2. What is the actual measurement, and what is its uncertainty?

A number without an error bar is not a measurement, it is an impression.

"The planet is 1.1 times the radius of Earth" means something quite different from "1.1 ± 0.4". The second is consistent with 0.7 and with 1.5 — which is the difference between a rocky world and something with a thick envelope. The uncertainty is not a technicality appended to the result. Very often it is the result.

3. How strong is the detection?

Astronomy and particle physics report detection strength in sigma — how unlikely the signal is if nothing were there, expressed in standard deviations. Physics conventionally reserves the word discovery for five sigma: the ATLAS announcement of the Higgs boson, for instance, reports its result as "a significance of 5.9 standard deviations" rather than as a probability [2].

The bar is set that high because lower-significance excesses are common and many of them do not survive more data — which is a statement about how the threshold is used, not something that paper argues.

Two rules of thumb that survive contact with reality:

• 3σ is interesting. 5σ is a discovery. A "tentative", "hint of" or "possible" detection is

usually a 2–3σ result, and a meaningful fraction of those evaporate with more data.

• Sigma measures one thing only: how unlikely the data are under the null hypothesis. It says

nothing about whether the interpretation is right. A five-sigma detection of a signal can sit beside a completely wrong explanation of what caused it.

4. Is the exciting word doing real work?

Three that almost never mean what a reader assumes:

"Earth-like" — usually means similar in radius, and nothing else. Not the temperature, not the atmosphere, not the surface, not whether it has one. The exoplanet archives [1] record radius and mass; "like Earth" in any richer sense is a claim about things that have generally not been measured.

"Habitable zone" — the orbital distance where liquid water could exist on a planet with a suitable atmosphere. It is a statement about orbits, not about habitability, and the boundaries move with the model: the widely used calculation of Kopparapu et al. puts the Sun's conservative zone at 0.99–1.70 AU, which places Earth near its inner edge [4]. Venus, at 0.723 AU, is well inside even the optimistic empirical limit — a planet can be the right size at the right distance and still be Venus.

"Signs of life" — in practice, a molecule detected that has a biological pathway and several non-biological ones. The abiotic explanations are in the paper. They are the first thing cut from the coverage.

5. What would make this wrong?

Good papers say. They name the systematic that would mimic the signal, the assumption the model rests on, the follow-up observation that would settle it.

If an announcement contains no such sentence anywhere, that absence is itself information — either about the result, or about who wrote the summary you are reading.

6. Who is telling you, and how many hands has it passed through?

An agency release about its own mission is authoritative on the mission and enthusiastic about the result. An outlet's article about the release has lost a qualifier. An aggregator's article about that article has lost another.

Follow the chain back to the paper. If the chain does not lead to one, that is the finding.

The short version

Ask what was measured, how precisely, how strongly, what else could explain it, and who checked. If an announcement answers all five, it is probably solid regardless of how modest it sounds. If it answers none of them but uses the word "breakthrough", the confidence is coming from somewhere other than the data.