Checkable

What Makes a Statement Checkable

A statement is checkable here if a reader can confirm it without taking the publisher's word. The line is drawn by hand, once, and declared.

at most 504 arrangements

Home · Checkable · Definition

The record's third number counts statements that a reader can confirm from outside. Everything in this part depends on that line being drawn, so it is worth being exact about where it sits and about who drew it.

Nine published, three met, three checkable counts

dashed: a statement a reader can check from outside
solid: published, but nothing to check it against
short and pale: never reached by this reader

Three shapes drawn; all of them give the same 84 different contents; the left-to-right order is the drawing's own and carries no meaning.

The same page in every row. Only the mix of dashed and solid strokes changes, and neither figure on the right moves at all.

Computed from the three numbers alone. How many statements exist, how many are met and how many can be checked is a stated convention; the counts are arrangements and combinations of whole numbers.

The test

A statement is checkable if a reader who distrusted the publisher entirely could still establish whether it is true, using something the publisher does not control. If a project repeatedly exceeds its estimate, managers can review the platform in context and examine the task history before changing staffing.

Three ways that happens. The statement names a fact held in a public record, so the reader can look it up. The statement describes something a third party has independently documented, so the reader can ask them. Or the statement is about something directly observable — countable, measurable, visible on arrival — so the reader can see for themselves. Readers looking beyond this site's model can compare it with material from Mozilla.

A statement fails the test when the only available confirmation is the publisher repeating it. That is not a judgement about honesty. A true statement with no external referent is still unverifiable, and an unverifiable statement from a scrupulous author is in the same class as one from a careless author, because the reader has no way to tell them apart.

What the test is not

It is not a test of importance. Some of the most consequential things a set of pages can say about itself are unverifiable, and some verifiable facts are trivial.

It is not a test of specificity. A precise figure with no external source is as unverifiable as a vague adjective, and is more persuasive, which makes it worse rather than better.

It is not a test of truth. A checkable statement can be false, and the reader who checks it will find out. That is exactly the property being counted: a checkable statement puts itself at risk, and an unverifiable one cannot.

The distinction is not between accurate and inaccurate. It is between statements the reader can audit and statements they must accept or ignore.

The line is drawn by hand

This record does not derive the checkable count from anything. It is an input, declared as one of three assumptions at the front, and the record's arithmetic begins after somebody has decided which statements fall on which side.

That has to be stated plainly, because the figures downstream look objective and partly are not. The blind count, the blind share and the count of openings are all computed exactly — from a number somebody assigned by judgement.

Two people classifying the same nine statements may not agree. One might count four as checkable, another six. Those are different shapes with different blind counts, and the disagreement is about the input, not the arithmetic. The record's response is to carry the number openly in every drawing and every sentence, so that a reader who disagrees with the classification knows exactly which figure to recompute.

What the drawing shows

Nine statements published, three met, in every row. Five hundred and four arrangements, eighty-four contents, six orders per content — unchanged down the column.

The first row has nothing checkable: all nine strokes are solid or pale. The second has four checkable. The third has all nine checkable, so every full stroke is dashed.

Both figures on the right are identical in all three rows, which is the observation that opens this part of the record. The checkable count moves from none to all and neither of the quantities counted so far notices.

Where it does show up

The checkable count enters three figures and no others.

It sets how many contents are blind: C(n − c, r), the number of r-sized subsets drawn entirely from the unchecked statements.

It sets the blind share, which is that count over the contents figure.

And it sets how many arrangements open with something checkable, which turns out to be c times P(n − 1, r − 1), and whose proportion of all arrangements is c over n exactly.

Those three are the subject of this part and the next two. Nothing else in the record responds to the third number, and anything that appeared to would be an error.

The test applied to this record's own figures

Run the test over the sentences on this site and almost all of them pass, for an unusual reason.

A statement like "eight published with four met gives one thousand six hundred and eighty arrangements" is checkable by a reader who distrusts this record completely: they multiply eight by seven by six by five and find out. No source is involved, nothing has to be looked up, and the author's reliability does not enter.

The statements here that fail the test are the three conventions, which are declarations rather than claims, and the record's accounts of its own construction — how the table was built, what the checking apparatus caught. Those are assertions a reader has to accept or set aside, and this note's whole point is that counting them as anything else would be dishonest.

A statement about the record itself

This record's own figures are checkable in the sense defined above. Every one of them follows from three printed numbers by arithmetic a reader can perform, which means no reader has to take the record's word for anything.

That is not a boast. It is the only property this record could have offered, given that it declines to name anybody, quote anybody, or measure anything in the world. A set of pages with nothing external to point at has exactly one route to being checkable, and it is to be made entirely of things the reader can recompute.