AquariumHQ
Search
evidence

When two sources are really one

AquariumHQ's corroboration figures fell this sprint, and nothing was removed. What changed is that the site started counting ORIGINS rather than registrations: it now resolves each citation to the upstream work it actually reaches, and where two registered sources terminate at the same one they count once. This guide explains what that means for the confidence figures on every record.

Short answer

Two databases publishing the same temperature range are not two pieces of evidence if both of them are quoting the same book. Independence is a property of where a figure originated, not of how many places repeat it.

The shape of the problem

Aquarium information has a small number of headwaters and a very large number of streams. A great deal of what circulates as separate sources — care sheets, databases, forum consensus, retailer copy — resolves, when followed, to a handful of twentieth-century compilations, and frequently to one of them.

That does not make the figures wrong. It makes a claim of corroboration wrong, and corroboration is what a confidence rating is largely made of. A number agreed by four sources reading one book is exactly as well evidenced as that book, and rating it higher is manufacturing certainty.

Two aggregator databases resolving to a single upstream workOn the left, two boxes labelled Database A and Database B, each of which a catalogue might register as a separate source. Arrows from both converge on a single box in the middle labelled 'one aquarium reference book'. A further arrow leads from that box to the right, to a box labelled 'one origin'. Below, a separate pair shows a field survey and a museum collection converging on nothing — two arrows leading to two distinct boxes labelled 'two origins'.Counted as two sourcesDatabase ADatabase Bone reference book1Also counted as two sourcesField surveyMuseum seriesits own measurementits own specimens2The catalogue now counts origins, not registrations.
Two databases agreeing is sometimes one measurement reported twice. Independence is a property of where a figure came from, not of how many places repeat it. Source: AquariumHQ source lineage model.

What this site found when it looked

Forty-four of a hundred and twenty-two attributed values in the catalogue's independence sample trace to aquarium keeping literature rather than to a measurement — overwhelmingly to one atlas, reached through a database that credits it. That is real evidence about how these fish are kept. It is not an independent measurement of what they need, and the difference had been invisible precisely because a database reported it.

The pattern is uneven by predicate, and the unevenness is informative. FishBase credits a publication for a species' temperature and its maximum length and credits nothing at all for its pH or general hardness on the same page. So the site's temperature evidence can be followed to an origin and a large majority of its pH and hardness evidence cannot, and the two were previously displayed as equally sourced.

What this means for a reader

  • A high confidence rating here now means several distinct origins, not several websites.
  • Where the origins collapse to one, the record says so rather than averaging it away.
  • A value with one origin is not thereby wrong. It is one measurement, and it should be read as one.
  • Concentration is reported as a finding in its own right. A single source supplying most of a catalogue's values is a risk even when every one of them is correct, because its silences and its defaults become the catalogue's too, and they do not announce themselves.