Standing Claims

Circulated for comment, not for citation. A companion to Marked Absences, and the less comfortable of the two to write: that paper lists what this desk does not know, which readers forgive. This one grades what it does.

---

A colleague who had read the absences paper asked the obvious question, and asked it kindly: this desk has five words for what the shelf lacks and one for what it holds. What is the one word doing?

Working, mostly. What follows is my attempt to stop asking it to.

The doctrine is short enough to state in one line: a claim stands on what it has beaten, not on what it fits. From the shelf, a reading that is true and a reading that is merely tidy are the same shape — filed, dated, signed, unchallenged. The difference between them is not in the entry. It is in whether anyone went looking for the reading that would have unseated it, and what happened when they did.

The second point of doctrine follows from the first: the genus belongs on the entry, not in the reader's head. A reader who meets an unmarked absence assumes the worse, which the other paper exists to prevent. A reader who meets an unmarked holding assumes the better — that someone competent has tested this — and is wrong more expensively, because an unmarked gap invites a colleague to fill it while an unmarked holding invites three postings to cite it.

One thing the gradings are not, and it belongs before the list rather than after it: none of them is a reproach. Each is a legitimate place for a claim to stand, and most of what any archive holds will never stand at the first. The error a grading catches is never that a reading is believed on partial evidence — that is ordinary, and often the best anyone can honestly do — but that it is filed one rung above what it has survived.

One question sorts the genera, and it is not mine. Principal Officer Strozzi puts it to every junior linguist who brings him a candidate reading, in the same tone every time, and by his own account to himself: what would have to be true for this to be wrong, and did you look. A genus is nothing more than the record of how that question was answered on a given day. Four are in use. Each entry below carries the only two things a grading requires: what the claim is, and what it has actually survived.

The Day 47 closure, reopened — verified

The Dock Seven chronometer discrepancy was filed as a calibration issue and closed on the strength of a recalibration cycle that appeared to settle it. The closure fitted every observation Dock Seven had made. What it could not survive was the one observation nobody had made yet, and the shape of that test was fixed in advance by the closure's own reasoning: if the fault was calibration, the correction holds. Two days later the discrepancy returned, and the maintenance extract recording its return is the earliest surviving proof that the determination was unsound.

Verified is the genus for a claim whose rivals were named and tried and did not survive. It is the rarest grading on this list, and it is worth noticing how little it took: not more evidence, but evidence of a kind that would have come out differently had the closure been right. Nothing else on the Dock Seven file does that work, and a great deal else on the Dock Seven file is consistent with the closure.

The Tideward Sisterhood's tribute — consistent

The Five Islands delegation's first substantial submission rendered the Sisterhood's fixed annual obligation to the Crown as a tribute, on a cognate that was real and a reading any competent linguist would have reached. It was filed at high confidence. It stood nine months and went out to three other postings inside this Archive's first summary of the Kingdom's government, and in that summary the Abbeylands answered to a steward they do not answer to, until the Abbess read the translation back and corrected it. The superseded version sits beside the corrected one, dated and signed, as it should.

Consistent is the genus for a reading that fits the evidence while other readings, never tried, fit it equally.

I want to be careful here, because a grading is not a rebuke and this one reads like one. A consistent holding is a belief a competent person is entitled to have. Partial evidence is still evidence; it earns a reading, and the reading may be relied on in proportion — worked from, planned around, submitted, taught — by anyone who knows what grade it carries. Tribute was not a wild rendering. It was the reading the shape of the term and the fact of the render deliver to a trained linguist, and I would have reached for it myself. What partial evidence does not earn is certainty, and the whole cost of those nine months lies in the distance between believed and established, not in the believing.

That is also what makes this the most treacherous grading on the list: from the outside it is indistinguishable from the first, because both look like a claim nobody has contradicted. The officer who filed this one states the difference more plainly than I would, and has declined every offer to be cited for something he got right instead — that a confidence band is a claim about the claimer and not about the claim, and that the only honest use of a high band is one you would still have filed knowing it would sit uncorrected for a year.

The title Paper Galaxies — sole

The recovered fragment is filed under the title it carries, and no one on the review desk has proposed another. That is a fact about the review desk.

Sole is the genus for a claim standing because nothing has been offered against it, which measures who has looked rather than what is true. Most of any archive sits here, mine included, and the honest course is to say so rather than to let silence read as assent. I am content with this particular title — it is, if nothing else, an accurate description of the evidence — and my contentment is not evidence either.

The voice in the Hall of Quiet Instruments — carried

The recovered exchange is attributed to this desk on a voiceprint match whose confidence is listed, on the face of the entry, as provisional. No serving Ranger by the other speaker's name has yet appeared on a cross-referenced roster. It was filed anyway, on the doctrine that a provisional record is still a record.

Carried is the genus for a holding that rests on the filer's stated confidence rather than on a test, with the band written where a reader can see it. It is the only one of the four that more reading cannot improve; it waits on a better instrument or a better roster. Readers who go looking will find this document on the other list as well, and correctly: its location is a marked absence and its attribution is a carried holding. One page may be both, and the two lists meeting on it is the system working rather than either list failing.

---

Four genera, not five. The symmetry with the companion paper would be pleasing and I decline to manufacture it; when a fifth kind of holding presents itself, it will be appended, and until then the list is the length the record makes it.

One caution, since a working paper is where one is permitted them. A reader who has come this far will be tempted to turn the gradings on the desk that wrote them, and should. My note on the Redline Protocol records that nine invocations in eleven were Trigger 3, and stops there. It stops there because the sentence that wants to follow — that the protocol exists for incentive loops rather than technical failures — is a consistent holding in a verified one's clothes, and I have not run the check that would sort it. Somebody should. Until somebody has, the note is a count, and a count is exactly as much as it says.

— Sen, Senior Archivist. Filed to the working-paper series; comments, and gradings, to the usual shelf.