Hidden, Lost and Forbidden Knowledge

This is an account of a working that failed in a specific and instructive way, written after the fact and stripped of the parts that would make it sound better. The aim was to assemble a dossier of hidden knowledge. The method was to read widely and record what seemed relevant. The failure was that the reading quietly accumulated around a story about protective custodians, and by the third month I was no longer able to tell which parts of the dossier were findings and which parts were the shape the story had imposed on them. What follows is the account, then the three silent failures, then the second dossier built the other way.

The working, and the story built into it

The aim, written in the imperative as a working aim should be: assemble a defensible dossier of the material I want to use. That is a good aim, specific enough to be finished, and I have written it out in full because the way it went wrong is visible in the wording. It says nothing about guardians, nothing about suppression, and nothing about who kept anything safe.

The story arrived in week two, from a passing remark in a piece of secondary commentary, and it was doing a lot of work immediately. The claim was that the tradition had been guarded, that things had been withheld deliberately, and that a small body of material was the surviving remnant of that custody. It is an attractive story. It explains the fragmentary nature of the record, it explains why the material feels like it has an owner, and it makes a scattered set of books into a coherent object. I accepted it without checking it and everything after that was shaped by it.

Month one, and the sources that fit

Month one was productive in the way this failure usually is. Four sources went into the folder, all of them serious, all of them relevant, and all of them more supportive of the guardian story than the evidence warranted. A fragmentary Jewish apocalyptic text in a category where one expects exclusions. A short modern pamphlet that everybody starts with. A book with no known author. A translation history that showed material moving and being combined.

None of these is false. That is what made the month feel like progress. The exclusions are documented, the pamphlet is what it claims to be, the anonymous book is genuinely unattributed, and the translation history is interesting. What I did not do was ask whether the pattern was mine or the sources'. Four relevant documents do not make a pattern, and a pattern that appears in the first month when you expected one is exactly the pattern you should be most suspicious of.

The month the fit stopped

The fit stopped with a source that was directly relevant and did not fit. A modern treatment of the same material that took no position on custody at all, and instead documented what could and could not be established about transmission. It was less exciting than anything else in the folder and it was the only thing in the folder that was actually evidence.

I read it, found it careful, and shelved it as overly cautious. The thing I noticed afterwards is that I had a category ready for it, which was too academic, and I used the category to avoid engaging with the content. That is a distinct failure from confirmation bias. Bias is preferring what you like. What happened here was classifying inconvenient material into a bin and leaving it there, and the bin was labelled in a way that made the dismissal look like judgement.

Guardians, and where the idea comes from

The guardian story is not crazy and it is not invented by me. It is a natural product of a fragmentary record. If material is missing, the missing has to have a cause, and the available causes are loss, suppression, or custody. Loss is unsatisfying because it implies carelessness. Suppression is dramatic. Custody is the one that makes the surviving material feel intentional rather than accidental, which is why it spreads.

The problem is that custody is a claim about intention, and intention leaves a document. If somebody held and protected something, there should be records of the holding, a list, a transfer, a lineage of holders. What there is instead is an absence, and the absence is being read as a positive. An absence cannot be evidence of custody. It is evidence of absence, and the two are not the same, and the whole dossier was resting on treating them as the same.

The primary text, read against the story

When I finally read the primary material properly rather than through the frame, the story came apart. What the fragments actually are is a body of apocalyptic literature in which some pieces were known to a few writers in one period and absent from later editions, and the interesting fact is the documented shift rather than any custody of it. That is a smaller claim and a better one, and it has the enormous advantage of being checkable.

Reading it against the story took one evening. The fragments do not refer to guardians. They do not refer to a custody. They refer to a subject matter that has nothing to do with the folder's thesis. The fit I had been feeling came entirely from the frame and not at all from the text, and once I had read the thing itself the frame had nowhere to attach to, which is an uncomfortable way to discover that your finding was an artefact of your procedure.

The pamphlet, which is not a guardian

The short modern pamphlet was the worst offender and the most obvious. It was printed, sold in shops, cheap, and freely available, and it is where most people in this field begin because it takes a weekend to finish. Nothing about it is guarded. Nothing about it was taken from anyone. It is an instruction sheet written in the late twentieth century by somebody who had already decided what the practice was for.

It had ended up in my folder as evidence of a custody tradition because it is old enough to feel like an artefact. The category error is that old was read as withheld, when old here means modern and available. Adding it to a dossier of guarded material was the single most damaging thing I did, because it made the folder look better sourced while making it worse. The pamphlet is a good document. It is a document of its own decade, and it is not evidence of anything older.

The anonymous book, which guards nothing

The unattributed book was the other structural problem. An unknown author with a doubtful lineage is a book whose provenance cannot be checked, and I had it in the folder doing the work of an early text. It is evidence of what someone was arguing in 1908. It is not evidence of what any earlier text said, and no amount of careful reading turns it into that.

The custodial frame made this worse rather than better, because the frame rewards exactly the documents that resist checking. A text with a known author can be placed in a network of earlier positions. A text without one cannot, so the frame treats its opacity as depth. It is a poor trade, and taking it is what made the folder feel like it contained something hidden when in fact it contained four documents and a shape I had supplied myself.

Three ways the practice fails quietly

The first is the frame arriving before the aim. The aim I wrote said nothing about guardians and the story was in the folder by week two, which means the aim was doing nothing. The second is the convenient bin. Material that did not fit was classified as overly academic and shelved, which felt like judgement and was actually avoidance. Neither produced a feeling of dishonesty and both damaged the work.

The third is the fourth column that never got filled. I had a column for things I looked for and did not find, and I wrote in it twice. The folder was, in effect, a list of confirmations with two exceptions noted, and a list of confirmations has no information content, since I would have produced it no matter what I had read. The three failures share a root, which is that nothing in the procedure was capable of telling me I was wrong, and I did not notice that until the second dossier failed to improve.

A second dossier, built the other way

The second attempt took three hours and went completely differently. I wrote the failing sentence first: the tradition is plural, custody cannot be documented, and nothing in my folder indicates deliberate concealment. Then I built the folder to test that sentence, which meant the fourth column was the main one and the guardian story had nowhere to go except as a thing I looked for and did not find.

I kept four documents and dropped the rest. Each one is now labelled with what it can support, and the folder reads as a modest set of findings about transmission rather than a body of guarded material. It is less dramatic and it is completely defensible, and every claim in it can be checked by somebody who does not trust me. The second dossier is the one I use. The first one is in a drawer and I have not thrown it away because the record of the failure is worth more than the folder was.

What the field notes show

Read over, the notes show something more specific than a general lesson about confirmation bias. They show that the frame did not distort the individual documents, every one of which I still stand behind, and that it distorted the relationship between them. Four true statements arranged under a false connecting claim produce a false finding, and the individual accuracy is what made it invisible.

They also show that the productive hour was the one reading the inconvenient modern treatment, which I had dismissed as overly cautious in month one and which turned out to be the only document in the folder doing any work. And they show that the moment the frame lost its anchor, most of what I had accumulated stopped being findings and became material. The interesting question is why I could not tell the difference at the time, and the answer is that the two look identical while the frame holds.

The claim you have to give up

Not the whole guardian idea, because it is a reasonable hypothesis about gaps in general. The specific claim you have to give up is that custody explains this particular record. It does not, because there is no documentation of it, and the documentation exists for plenty of other things, which makes its absence here meaningful.

The claim is worth giving up early because it is load-bearing for a lot of downstream reasoning. If the material was guarded, then a small amount of material is significant and knowing it is valuable. If the tradition is plural and transmission was ordinary, then the same material is a sample of a much larger body, most of which is lost, and knowing the sample tells you much less. Both readings produce a practice. Only one of them lets you be wrong about what you are doing.

Three lines, and the one of them that transfers

Three lines, and they are the same three I now use. What I claimed at the start, in a sentence someone could check. What the record actually contains, including the two entries in the fourth column. And what I have to give up, which here was the custody claim and, more usefully, the assumption that a folder which feels like it contains something is thereby doing work.

The last of those is the transferable part of this failure. A dossier that feels substantial and a dossier that is reliable are different objects, and the feeling is generated by the frame rather than by the contents. If you take one procedure from this account, take the order. Failing sentence first, record open in advance, fourth column as the main one, and everything you look for and do not find written down where you can see it.

Frequently Asked Questions

Is the idea that the tradition was guarded ever defensible?

As a general explanation of why records are missing, yes, and it is a natural one to reach for. As a claim about this specific body of material, no, because custody leaves documentation and there is none here, which is informative given that plenty of other things are documented. The honest position is that transmission was ordinary and much is lost, which is a less flattering claim and the one that can be checked.

How do I know if my own dossier has a story in it?

Write the failing sentence first, before you read anything, and see whether your sources can contradict it. If you cannot write one, your dossier is a list of confirmations. The second check is whether you have a bin for inconvenient material, because a labelled bin is where a frame protects itself. Most dossiers with a story in them also have a bin, and both features appear at about the same time.

What should I do with a dossier that turned out to be shaped by an assumption?

Keep it and date it. The failure record is more useful than the folder, because it tells you which procedure step failed and at which week, and that is reusable. Then build the second version with the order reversed, which takes a few hours rather than months. Mine was four documents and one false connecting claim, and the second version is the one I actually use.

Does any of this mean the material is not worth studying?

No, and the opposite. This is the reason to study it, since the fragmentary record, the documented exclusions, and the plural transmission are the actual findings, and they are more interesting than a secret. What is not worth studying is the version of the record where someone was guarding something, because that version has to be invented and every element of it has to be supplied by the reader.

How many sources make a dossier worth keeping?

Three is enough to start, and what makes it worth keeping is not the number but the labelling. Each document should carry a line saying what kind of claim it supports, and one saying what it cannot be used for. A folder of four labelled documents that supports a modest claim will teach you more than a shelf of twenty that supports a large one, and it is the labelling rather than the reading that does the work.