How we work
Everything here is sourced, and connected to what it comes from and what it becomes. What follows is how we decide what to say, and how sure of it we are.
Two questions, kept apart
Every figure on this site is shown with the source it came from and with two judgements about it, which look alike and are not. Keeping them apart is the most useful thing on this page.
They are evidence quality and confidence. Evidence quality asks how good the material we are reasoning from is. Confidence asks how much you should trust the conclusion we drew from it. A claim can rest on two first-rate sources and still warrant only medium confidence, because those two sources disagree.
Confidence is per claim, never per page
You will not find a confidence score for quartz. You will find one for “quartz has a Mohs hardness of 7” and a different one for “quartz was first described under that name in the sixteenth century”. Those rest on entirely different evidence, and averaging them would produce a number that describes neither.
Where sources disagree, we say so
Reference works genuinely disagree about mineral hardness ranges, alloy composition bounds, discovery dates and type localities. The tempting response is to pick the most authoritative figure and present it cleanly. We do not do that, because a single number presented cleanly is a claim that the disagreement does not exist.
Where measured values legitimately span a range, we show the range rather than a midpoint. A midpoint is a number nobody measured.
Scales that cannot be subtracted
Mohs hardness is an ordinal scale. Quartz at 7 is harder than calcite at 3, but it is not “4 harder”, and it is not “2.3 times harder” — those quantities do not exist. So when we compare two things, we say which kind of scale you are looking at — one you can do arithmetic on, or one you can only put in order.
What you will not find here
A figure that looks wrong, and anything bearing on safety, are absent rather than present with a caveat attached. So an absence here carries meaning: if a hazard figure is missing, it is missing because it could not be stood behind, and a safety data sheet is the place to look instead.
Licences, and what “unknown” means
For every source, its own published terms about reuse are set out on the data sources page. Where those terms could not be found, they are shown as unknown — which means nobody has established them, not that reuse is permitted. Where a source stands affects what you see: a figure here is always one we are free to quote, whatever else was read along the way.
The same applies to photographs, which is why there are none here yet: a picture whose origin and rights cannot be established is a picture we cannot honestly show you the provenance of, and provenance is the point.
A longer page is not a better one
Some pages here say a great deal and some say very little, and the length tells you nothing about how sound the contents are. A long page can rest on weak evidence and a short one on excellent evidence. Which it is, is shown on each figure rather than summarised for the page as a whole — so read the evidence beside the number you care about, not the size of the page it sits on. Coverage says which subjects are thinly covered.