The Value Not Yet Asked For

I.The First Known Thing — the decision to destroy

archival science · appraisal

An institution of any size produces more paper than anyone can keep. The archivist's central professional act is therefore not preservation but selection: deciding which small fraction of what an organisation generates will be transferred to an archive and which will be destroyed. The discipline calls this appraisal, and it has argued about it for a century. Hilary Jenkinson, writing in 1922, held that the archivist must not do it at all — that a record's value as evidence rests on nobody having chosen it for its interest, so the creating body should decide what to destroy on its own administrative grounds and the archivist's duty begins afterwards, with custody. Theodore Schellenberg, facing the volume of the American federal government thirty years later, took the decision back and gave it a vocabulary that is still the working one. A record has primary value, its use to the agency that made it — administrative, fiscal, legal, operational — and it has secondary value, its use to everyone else once the agency has finished with it. Secondary value he split in two: evidential value, what the record shows about the organisation and functioning of the body itself, and informational value, what it happens to contain about persons, places and phenomena, judged on uniqueness, on form, and on importance to future research. Retention schedules enact the scheme: a class of records, a period, and then transfer or destruction, applied to material nobody reads first.

the asymmetry sitting in the middle of the method

The two values are not two halves of one quantity. Primary value is measurable on the day the decision is taken, by asking the office that made the record whether it still needs it. Secondary value is a forecast about questions that have not been asked, by researchers who are not yet working, in fields that may not yet exist — and the history of the profession is largely the history of that forecast going wrong in one direction. The turn towards social history from the nineteen-sixties made scholarly gold of exactly the material that earlier appraisers had destroyed as routine: case files, tax and poor-law paperwork, registration returns, the granular clerical residue of ordinary lives, discarded precisely because it was repetitive and had no evidential value for the institution. Hans Booms argued that the values of a society rather than the judgement of an archivist should form its documentary heritage, and that those values move. F. Gerald Ham, in The Archival Edge in 1975, put the objection to the profession plainly: left alone, archivists build a record that mirrors the research fashions of their own moment, and the mirror is not noticed because nothing outside it survives to disagree.

what the destroyed record cannot do

That last clause is the structural point, and it is not the same as saying the forecast is hard. Consider what evidence an appraisal rule can ever receive about itself. A record kept and never consulted produces a visible cost and a visible reproach — storage paid for, shelves filled, generosity demonstrated. A record destroyed that would have answered a question asked in 2031 produces nothing at all: the researcher meets an absence, and an absence is indistinguishable from a matter that was never documented in the first place, or documented somewhere else, or documented in a hand nobody can read. The rule is therefore audited on one side only. Its errors of retention are counted and its errors of destruction are not counted, not because anyone is careless but because the counting instrument was the thing destroyed. The profession has a name for the resulting shape of the record — archival silences, the missing, the lost and the never-created — and a body of work on how they form and what they conceal. Correction, when it comes, comes from outside the archive: an inquiry, a commission, a scandal, someone with standing to name the specific file they needed and to make the absence itself a public fact. Terry Cook's macro-appraisal is the methodological answer to all of this, and its move is the one worth carrying: stop appraising records. Appraise functions. Decide which activities of the institution, and which of its points of contact with the society around it, must be documented at all; then keep whatever those functions generate, whatever it looks like. The unit of the decision shifts from the document to the thing the document would be evidence of, because that is the part a person in the present can reason about.

II.The Second Known Thing — the list of tissues

drug discovery · toxicity and safety pharmacology

Walk to a hall where nothing is destroyed on purpose. Before a molecule may be given to a person, its sponsor must characterise what it does to animals that have nothing wrong with them. Two traditions do this work. Repeat-dose toxicology runs the compound in a rodent and a non-rodent species for weeks to months at several multiples of the intended exposure, with body weights, food consumption, clinical observations, haematology and clinical chemistry through life, and at the end a necropsy: organ weights, then tissues taken, fixed, embedded, sectioned, stained and read by a pathologist. Safety pharmacology, codified in the ICH S7A guideline in 2000, asks a narrower and more acute question — what the compound does to the functions whose sudden failure kills — and answers it with a core battery of three: the central nervous system, the cardiovascular system and the respiratory system, with follow-up and supplemental studies where something gives cause.

what is collected is fixed in advance, and by a list

In both traditions the content of the collection is settled before the study begins, and settled largely by convention rather than by the compound. The Society of Toxicologic Pathology's recommended tissue list, published by Bregman and colleagues in 2003, sets out the tissues to be taken and examined in repeat-dose and carcinogenicity work; guidelines and house standards fix the clinical pathology panel; S7A fixes the three vital systems of the battery. These are good lists, assembled honestly, and each item is on them because it has produced findings. They are also under steady downward pressure, from cost, from study duration and from the commitment to reduce animal use, so the live question the field asks about the list is whether anything currently on it can come off — a question about primary value, asked of the study that is being run, answered from the study that is being run.

and the archive nobody appraises

The material does not then vanish. Under good laboratory practice the specimens and raw data are archived: wet tissues, paraffin blocks, stained slides, the study records, retained plasma and serum, held for years under conditions that are themselves documented. Blocks can be re-cut and slides re-read a decade later; a retained sample can be re-assayed with a method that did not exist when it was drawn. And this is exactly what happens when a signal arrives late. An enzyme elevation in a phase II cohort, an unexplained event after approval, a competitor's compound failing on an organ nobody was watching — and the sponsor goes back to its own archive and asks it a question it was never designed to answer. The value of that reuse is well understood in the aggregate: the eTOX project pooled legacy preclinical safety studies from more than a dozen companies specifically to answer questions the original studies had not posed. What does not exist anywhere in the discipline is a measure of how often the archive answers, or any moment in the design of a package at which its ability to answer a question not yet asked is scored, costed, or traded against anything else.

III.The Bridge

no tissue list cites Schellenberg

The halls share nothing. Archival science and toxicologic pathology have no common citations and no review spans them; the only word they hold in common is archive, which in one hall names a century-old discipline about selection and in the other names a room with a temperature log and a retention period. What they do have in common is the shape of a decision. Both must fix, in advance and irreversibly, what will be in a position to testify later; both must take that decision using only the value the material has to the people taking it; and in both the material that is not collected is silent afterwards in a way that is indistinguishable from there having been nothing to collect. One of the two halls has spent a hundred years discovering that this is the hard part, and has a name for the structure and a method that answers it.

X: the rule is graded only by what it spared

Call it appraisal asymmetry. Where what will be kept has to be chosen before the questions are known, the choice can be made only on the value the material has to its maker now, and that value is not the value a later question will give it; and because what was not kept cannot afterwards report what it would have answered, the rule that made the choice receives evidence from its survivors alone, so it is confirmed by its own output and can be revised only by a shock from outside it large enough to name the record it did not keep. From which follows the half the safety hall has not ported: a collection appraised entirely on its value to the study that generated it has not concluded that its value to later questions is low — it has had no occasion to form the judgement, and the adequacy it appears to demonstrate afterwards is produced by the same rule that set it. It is worth being exact about what is and is not new here. That a rule is confirmed by its survivors is a familiar thing, and on its own it is only a caution. The archival hall's contribution is the two parts that are not familiar: first, that the value available at the moment of decision and the value that will be wanted are not merely differently distributed but need not be related at all, so improving the first buys none of the second and a better-run study is not thereby a more answerable one; and second, macro-appraisal's move, which converts an unforecastable quantity into a forecastable one by changing the unit. One cannot predict which question will be asked of a safety package in 2033. One can state which physiological functions the package leaves able to testify at all — and that is a property of the protocol, readable before the first animal is dosed.

IV.The Question That Becomes Possible

neither discipline would ask this alone

Every nonclinical safety package carries numbers. It has a cost, a duration, an animal count, a compliance status, a count of findings and a set of margins derived from them. None of them has a secondary yield: the proportion of the later safety questions put to a package that its retained material and its recorded endpoints turn out to be able to answer. The archival hall says that quantity exists, that it is the one that matters for the whole life of a compound rather than for the decision that generated the package, and that nothing in the way a package is currently designed or judged is an estimate of it.

the appraisal-asymmetry hypothesis

Does a nonclinical safety package have a secondary yield — the proportion of later safety questions put to it that its retained material and recorded endpoints can actually answer — that is uncorrelated with every measure of its primary quality, and can that yield be predicted before the study is run from which physiological functions the protocol leaves able to testify, rather than from which tissues it lists?

why it would matter

Because the two orderings would come apart in a place that is cheap to fix and currently invisible. A tissue can be on the list, taken, fixed and read, and still leave a function unable to testify: an organ sampled morphologically but never functionally, so a question about what it was doing rather than what it looked like meets a slide and no measurement; plasma retained abundantly but at times chosen for the exposure question of the day, so a later kinetic question meets a sampling schedule with a hole in exactly the phase it needs; telemetry recorded over a window that closes before a delayed effect would appear, so the trace exists and ends. None of these is a failure of the study as designed, and none of them would show up in any audit of it. The useful consequence is not that packages should collect more — the constraint on animal use runs the other way, and the archival hall's own answer to a volume constraint is to re-appraise rather than to keep everything — but that a package could be made more answerable at the same cost by moving a sampling time, retaining a fraction differently, or scoring a function that is already being observed. That trade is not available to anyone until the yield is a number, because until then there is nothing on the other side of it.

V.How It Could Be Proven — or Broken

a candidate must be breakable, or it is merely poetry

An intercognix candidate earns nothing until it names its own test. This one has an unusual advantage: the interrogations it needs already happen, routinely and on the record, and nobody has counted them.

the honest status

Note the narrowness. The claim is about small-molecule nonclinical safety packages and what they collect, measure and retain. It proposes no new assay, no additional tissue, no revision of the recommended list, and no suggestion that a larger package is a better one. It does not say the standard lists are wrong; the lists are where findings are found, and that is a real defence of them. It says only that they have never been scored on the axis the archival hall names, that scoring them on it requires moving the unit from the tissue to the function, and that the record needed to try is one the industry already generates and does not keep as a record. A caution about what the analogy does not carry: an archival silence is very often a question about whose life went undocumented and by whose decision, and the profession's argument about appraisal is a moral argument as much as a methodological one. Nothing here borrows that. What is borrowed is structural — that a selection rule which must act before the question is asked can be graded only by what it spared, and that the way out is to appraise the function rather than the record. Until a portfolio's interrogations have been counted and a coverage score has been tested out of sample against them, this remains exactly what its stamp says: a candidate. A structural echo between two halls of the library, waiting for someone with the tools of both to walk the corridor.