A piece of writing can be accurate, well organized, and carefully worded, and still fail with the person it was written for. When that happens the writer is usually told something general, that the piece was heavy going or hard to follow or somehow not quite right, and the report is honest but it does not say what to change. This paper is about the layer of a document where that kind of trouble lives, and about taking that layer apart into pieces small enough to adjust one at a time.
(c) 2026 George Georgalis <george@iuxta.com> Unlimited use with attribution.
Begin with an ordinary situation. A writer sends a document to a colleague who needs it, and the colleague comes back and says it was too complicated. The writer looks at the draft again and finds nothing that is actually wrong. The facts are right, the order is sensible, the sentences parse. Every candidate repair looks like a way of throwing something away, and the writer has no way to tell which of the things on offer is the one that was causing the trouble.
The complaint was accurate. What the colleague experienced was real, and the description was as good as such reports ever get, which is to say it named the feeling rather than the cause. Difficulty of this kind does not arrive with a label attached. The reader notices that the reading was hard and does not notice, because there is nothing to notice, which particular property of the prose made it hard.
That gap between an accurate report and an actionable one is the problem this paper addresses. The report says the document failed. It does not say which adjustment would have prevented the failure. What follows is a way of listing the adjustments in advance, so that a report of that kind can be resolved into a specific change instead of a general worry.
Something has to be separated before any of this becomes tractable, and it is the separation most writers already make without naming: what the document says, as against how it reaches the person reading it. Call the second one delivery. Delivery is the manner in which finished content arrives at a particular reader, and the reason it is worth naming separately is that delivery can be changed while the content stays exactly as it was.
This is easier to see with an example than to accept as a claim. Take one idea, that a compressed sentence and a distributed explanation carry the same information at different costs per moment, and write it two ways. The first way is the sentence you have just read. The second way takes several sentences: a writer can put an idea into a single tight clause, where the reader must assemble all of it at once; or the writer can lay the same idea out across a paragraph, where each sentence adds one part and the reader assembles it gradually; the total is the same, and the demand made at any single moment is not. Both versions contain the same idea. Neither has lost anything. They differ in delivery, and a reader who found the first version hard would find the second version easy without having been told anything new.
The observation is old. Francis Bacon, writing in 1605, faulted Aristotle for treating the devices of rhetoric as serving proof alone, when their greater service is impression: forms that are equal in what they signify are unequal in how they strike, as a sharp and a flat blow of the same strength differ in what they pierce. He added that the proofs of logic are the same for everyone while the persuasions of rhetoric must differ according to who is listening, so that the same matter spoken to several people should be spoken to each of them differently.[^1] Content fixed, delivery variable, and the reader as the thing delivery answers to. That is the whole of the position this paper occupies, four centuries later and with the parts enumerated.
Enumerating them is the work. Delivery is not one adjustment but a number of them, and a writer who cannot see them separately cannot move one without disturbing the others.
Try enumerating them and the list gets long quickly, because a draft admits many adjustments and every one of them is real. The viewpoint can hold still or move. Sentences can open on ground the previous sentence prepared, or open cold. Contrasts can be packed into one clause or developed across a paragraph. New ideas can arrive together or one at a time. Terms can be substituted for plainer ones or kept and taught. The main message can sit at the front or at the end. The tone can stay level or rise.
Set that list beside a real reader and something becomes visible that the list itself conceals. The adjustments do not vary freely. A reader who cannot carry a reorientation in the middle of a section is, nearly always, the same reader who cannot carry a tightly packed sentence, and those are not two independent facts that happen to agree. They are one fact about a tired reader, arriving at the writer in two disguises and asking to be settled twice.
Whatever produces the need for a viewpoint that stays put tends to produce the need for a contrast that is spread out, at the same time, from the same source, in the same proportion. The source is the reader. Describe the reader precisely enough and the adjustments follow, which means the long list was never the thing to reason about.
The description is called a posture: the stable combination of circumstances a particular reader brings to a document. A posture is not a personality and not a profession. Two engineers of equal skill have the same profession and, on a given afternoon, may have entirely different postures, because one of them slept and the other did not. Posture is what is true of this reader, for this document, this week.
Six things are enough to describe it.
Fluency is what the reader already holds in this particular subject: the vocabulary, the concepts, the things that need not be explained because they are already there.
Stamina is the rate at which the reader can assemble meaning right now. It covers fatigue, time pressure, and position in the working day, and it is separate from fluency and frequently opposed to it, since a specialist at the end of a hard week holds everything and can assemble nothing.
Appetite is desire for compression, as against capacity for it. Some readers want a tight formulation and are irritated by a patient explanation of something they grasped in the first clause, experiencing the patience as their time being spent without their consent. Others will absorb difficulty when the material requires it and take no pleasure in the exercise. Appetite is the component most often assumed rather than assessed.
Scrutiny is how far the reader is evaluating the writer alongside the content, running from settled, through guarded, to adversarial by role. Scrutiny changes how ordinary devices are received: a compressed formulation that reads to a settled reader as efficient reads to a scrutinizing one as a claim to authority that has not yet been granted, and it does so regardless of how easily the sentence parses.
Triage is what the reader does with the document physically, whether reading it through, scanning the headings and descending selectively, reading the opening and deciding, or meeting one sentence in passing. Triage varies independently of everything else, since scanning is a habit of workload that expert and settled readers practice as readily as anyone.
Obligation is how far the reader is obliged to read at all. At the high end the attention is secured by role or relationship. At the low end the reader is present by choice and will leave the moment the prose asks for effort, quietly and without complaint, no obligation having been incurred.
Six assessments, each of which can be made by attending to one person, and none of which requires a theory of writing to make.
Six components are more than anyone wants to hold at once, and three of them collapse into a single quantity that does most of the work. Headroom is how much difficulty this reader takes before the prose starts to distort.
Headroom is bounded by the weaker of fluency and stamina rather than by their average, which is the specialist-at-the-end-of-the-week case stated as arithmetic. Holding every term is no help if nothing can be assembled, and being rested is no help if the terms are unfamiliar. Whichever of the two is lower sets the budget, and appetite then decides how much of that budget the reader actually wants spent.
The remaining three components stay outside headroom deliberately. Scrutiny is a stance, triage is a behavior, and obligation is a set of terms; none of them is capacity, and each of them governs adjustments that capacity does not reach.
One consequence has to be stated once, plainly, because it inverts what intuition expects. Headroom runs opposite to the settings that describe the prose. The more a reader can carry, the more the document may ask, and asking is the demanding end of every scale. So a reader with a lot of headroom produces low numbers throughout the configuration. This inversion is carried inside the framework so that the writer never has to perform it, but a reader of a configuration should know it is there.
An adjustment, once named and given a range, is a setting. The framework holds sixteen of them.[^2] A configuration is a complete set of values for a document.
Two conventions make a configuration readable without a manual.
The first is direction. Every scale runs the same way: 0.0 is the compact, mobile, forceful end, and 1.0 is the steady, gradual, low-pressure end. Put more usefully, 0.0 is heavy on the reader and 1.0 is light. Once that is known, a number can be read anywhere in the framework without asking which way it points.
The second is naming. Every setting is named for the quality that is at its maximum at 1.0. The setting that governs how long a viewpoint holds is called settled, because at 1.0 the prose is settled. The setting that governs how thinly new ideas arrive is called sparse, because at 1.0 they are sparse. A configuration then reads as a description of the prose rather than as a list of codes, and this is the entire reason the rule exists.
The rule cost something to follow. The obvious name for the force of a request is authority, and it was rejected, because authority is at its maximum where the setting is at its lowest and the name would have pointed the wrong way. The setting is called invite instead, which names the same relation from the side that grows toward 1.0. A rule that is never inconvenient is not doing anything.
So far the settings have been a list. They group into three instruments, and the grouping is not filing. Each instrument owns one kind of cost the reader pays, and the kinds are genuinely different from one another.
Load is the cost of understanding per instant, which is not the same as the amount to be understood in total. A short heavy conductor and a long light one carry the same current. The same material, differently distributed, arrives comfortably or does not, and the distinction between how much and how fast is the one thing about difficulty that generalizes.
Land is the cost of attention, and what survives the reading. A document that is emphatic everywhere emphasizes nothing, because emphasis comes out of uniformity: a document that is level anywhere, at any pressure, and then has one marked event, emphasizes exactly that event. This instrument decides where the marked event goes and what it is.[^3]
Vantage is the cost of knowing where a statement stands: who is speaking, from what position, at what remove, and as of when. Every shift of viewpoint asks the reader to locate the speaker again, and every layer of framing inside a sentence holds an anchor open until the sentence closes. This cost is the quiet one. A reader who has lost track of who is speaking does not report a difficulty with viewpoint; the reader reports that the writing was hard to follow, or simply stops.
Three costs, three instruments, and one posture that sets all three at once. They are given in the order a configuration prints them, which is the order a writer confirms them in: load first, because it is what a reader's complaint most often points at; land second, because its decisions are made before drafting begins; vantage last, because its costs are the silent ones and it is what remains to check when the other two do not explain the reception. The writer assesses the reader; the framework resolves the assessment into a configuration; the instruments compose or revise prose to occupy it.
Here is where a reasonable person usually objects. Spread a contrast out and surely something gets dropped along the way. It does not, and seeing why is what makes the whole framework usable rather than merely descriptive.
Take a compressed formulation and unpack it. Find its two poles, give each of them its own predication, let the contrast emerge from their adjacency rather than from their collision, and then spend the emphasis you have recovered on ordering, so that the pole which decides the matter arrives last. Every term of the original survives. What changed is where the precision sits, not how much of it there is.
The operation runs in the other direction with equal fidelity. Collapse a developed contrast into a single predication, drop the connective, and let the juxtaposition carry the logic that the connective was carrying. A reference card is built by running the operation backward on purpose, and it loses nothing either.
Since the content survives in both directions, something else must absorb the difference, and what absorbs it is length. Content, compression, and length form a triple of which any two may be chosen, and the third follows. A document required to be short, plain, and complete is not badly configured; it is over-specified, and the correct response is to say so rather than to attempt it. One declaration names which of the three yields, and the honest options are that length floats while the content is preserved, or that the length holds and something must be cut, which returns an editorial decision to the writer instead of letting compression absorb it silently.
One more distinction, and it is the one that decides what the framework can actually do. A setting names a region the prose should occupy. It does not name a limit the prose must not exceed.
A limit is satisfied by avoidance. It yields nothing under conflict, and its failure is silent, because a construction that was wanted and refused leaves no trace of having been either wanted or refused. A region is satisfied by occupation, and it yields usable information under conflict, because it declares not only where the prose belongs but how far it may wander and which way to move when something has to give.
So every setting resolves to three values rather than one. The center is the position the prose should occupy. The tolerance is the band around it within which the prose is still correct, and this band is also the experimental range: a first draft takes the center, a second takes the edge of the band that the first draft's reception points toward, and a third is worked out from the measured difference between them instead of guessed at again.[^5] The yield is the direction the setting moves when a conflict forces it off center.
Yield runs the same way almost everywhere, toward the light end, and the reason is an asymmetry in how the two failures behave. Asking too much of a reader fails silently and permanently: the reader experiences difficulty without being able to name its source, reports the writing as poor, and does not come back. Asking too little fails visibly and recoverably: the reader is mildly patronized, says so, and a second document corrects it. Given an uncertain assessment, the recoverable failure is the one to choose.
The transform is driven by the reader with four exceptions, and each exception is a fact about the document rather than about who reads it. No assessment of an audience, however careful, produces any of them.
The first is what kind of document this is. The same reader takes a different rhythm for a catalog than for an argument, and nothing about the reader distinguishes the two.
The second is whether the situation is a genuine emergency. Steadiness applied to a real emergency reads as negligence, and the reader cannot be inspected to find this out.
The third is whether the length was ever the writer's to choose, since a form, a limit, or a house standard may have fixed it from outside the reading entirely.
The fourth arrived last and is the least familiar, so it gets its own section.
The fourth question is whether the document is gathering facts for what comes next, or weighing facts against what was already done. The declaration that answers it is called office, and it takes one of two values. Under construct, the document treats its facts as inventory: the question is what is now true and what follows from it, and the order in which the facts are presented is largely the writer's to choose. A Under judicial, it treats them as evidence about decisions already taken: the question is not what is reasonable altogether but what was reasonable at each point in time.
This is the least familiar of the four and the one with the sharpest consequence, because sequence is the one part of delivery that reaches through to meaning. Three different orders run through any document that reports events, and only one of them belongs to the writer.
Chronology is the order in which things happened. It is usually not given. Sometimes it can be established from a record, sometimes it can be inferred with stated confidence, and often neither, in which case the honest report is that the order is not known.
Knowledge sequence is the order in which things became known, and it comes apart from chronology as a matter of course rather than as an oddity. A flaw introduced in March, exploited in June, and discovered in August has one chronology and a different knowledge sequence, and any judgment of what someone should have done in July turns entirely on the second.
Composition sequence is the order the document delivers. It is the only one of the three the writer controls outright, which is exactly what makes it the instrument through which the other two are either honored or overwritten.
Two settings govern the alignment. sequent sets how closely the composition sequence tracks the chronology. iuxta sets how far each claim arrives carrying the state of knowledge it belonged to, so that at its high end a claim stands according to its own moment and a moment that cannot be established is declared unknown rather than supplied. The two vary independently, and the pair that shows it is common enough to be worth recognizing: a bare incident timeline runs in perfect occurrence order while attaching no knowledge state to anything, and a finding grouped by subject may date every observation while following no order at all.
Of those two, the timeline is the dangerous one. It looks rigorous, because its ordering discipline is visible on the page, and its silence about what anyone knew at each point is not visible at all. A reader given facts with no moment attached to them supplies the only moment available, which is the reader's own present, and reaches a verdict on knowledge the actor did not have. The effect has been measured: people who know how something turned out cannot recover the judgment they would have made without that knowledge, and they do not notice that they cannot, so they overestimate what they would have known and what others actually did know.[^4] A document delivered without epoch marking does not merely fail to prevent this. It supplies the conditions for it.
Bacon located the mechanism in the passage already cited, and located it as the office of rhetoric rather than as an abuse of it. The affections, in that account, behold only the present, while reason beholds the future and the sum of time; the present fills the imagination more, so reason is ordinarily beaten, and eloquence wins reason's case by making what is future and remote arrive as though it were present. Moving a fact along the axis of time, relative to where the reader is standing, is how persuasion works when it is working honestly. Bacon then noted that the doctrines of contraries are the same and only their use is opposite, which is the entire argument for why this one declaration had to come from the writer. The same operation that spares a reader unnecessary history in the building case fabricates a false verdict in the judging case, and nothing in the finished prose distinguishes the two. No assessment of a reader tells the writer which one is being written.
One further property of these two settings is worth stating, because it runs against the habit the other fourteen build. Every other setting is satisfied by doing something to the prose. These two, at their high end, can be satisfied by an admission: an order that cannot be established is reported as not established, and an epoch that cannot be established is declared unknown. Neither permits the gap to be filled by inference. A supplied order presented as fact is the costliest mistake available here, precisely because it arrives wearing the appearance of care.
Now the complaint the paper opened with can be answered. With the settings named, a reader's report stops being a general verdict and becomes a pointer.
"I had to read that twice," said of a passage whose facts are correct, usually means understanding arrived faster than it could be assembled, and the repair is to unpack the dense clause into successive predications while keeping every term.
"Hard to follow," said of a document whose individual sentences are each perfectly clear, usually means the viewpoint moved more often than the reader could carry. The reader was working to reorient rather than to understand, and that cost is felt without being available to name, which is why the symptom gets described and the cause does not.
"I lost the thread," said of simple sentences, usually means new ideas were arriving faster than earlier ones were being closed out, so obligations accumulated in parallel instead of discharging in turn.
"Why are you defending that?" usually means the prose was arguing with an objection this reader had never raised, and a defense attached to an accepted claim raises the doubt it was meant to forestall.
"So the takeaway is," followed by the wrong phrase, usually means some subsidiary wording recurred more often than the message did. Whatever recurs is what the reader retains, and the mechanism operates on whatever happens to recur rather than on whatever was meant to.
Each of these corrects one document. Behind each of them stands a component of the posture that was assessed wrongly, and correcting that corrects every document for that reader thereafter. A passage that was too compressed points back at stamina assessed too high, which is the most common error of all, since writers under pressure compress while readers under pressure receive compression badly, and the two states often coincide inside a single exchange.
Several things do not move into the framework, and saying which ones is part of describing it accurately.
Accuracy stays with the writer. A well-configured document delivers a wrong finding efficiently, and no setting registers anything amiss.
Standing stays with the writer. A firm request presupposes an authority held independently of the document, so configuring one expresses a right rather than conferring it.
The judgment of whether the document was worth writing stays there too, and a perfectly landed message that should never have been sent costs more than a clumsy one that belonged.
Diagnosis stays with the writer, and this is the boundary that matters most. Every input is an assessment of one reader in one week. A confident transform applied to a wrong posture configures with real precision for an audience that is somewhere else, and the precision compounds the mismatch instead of exposing it.
Return to the colleague who said the document was too complicated.
The complaint was accurate and it named the only thing available to name. What produced it was one of a small number of conditions, each distinct and each with its own repair: understanding arriving faster than it could be assembled, a viewpoint that moved before the reader had settled into it, ideas opening faster than they closed, a message landing somewhere other than where the attention was resting. Every one of those repairs keeps the content entire, which is the useful part, because the content was carrying the value the whole time.
The framework's real economy is not that it lists the adjustments. It is that the adjustments turn out to answer to a single cause. A reader who cannot hold a viewpoint and a reader who cannot hold a compressed sentence are one reader, and the many decisions that felt like a balancing act were one assessment wearing many costumes. Given a name, the balance stops being a balance. Six facts about the person reading determine nearly all of the settings, and what remains is only what the reader was never going to answer: what kind of document this is, whether the situation is urgent, whether the length was ever the writer's to choose, and whether the facts are being gathered for what comes next or weighed against what was already done.
Which suggests where the effort is best spent, and it is not on the prose. It is on the reader the document is written to reach: what does this reader already hold, what can be assembled today, what is wanted, what is being weighed, how much of it gets read, and is the reading owed at all. The answers are usually obtainable by paying attention, and occasionally by asking. A writer who has settled that much before the first sentence will find that a good many of the decisions afterward have already been made --- which is worth something on an ordinary day, and worth a great deal on the day a draft comes back and the only thing anyone can say about it is that it was somehow hard to read.
Francis Bacon, The Advancement of Learning (1605), Book II, section xviii. The passages drawn on here are the account of rhetoric's office in paragraph 4, the remark on contraries in paragraph 5, and the criticism of Aristotle's colours in paragraph 6. Text at <http://socrates.acadiau.ca/courses/engl/rcunningham/1103/Bacon.htm>
The full set, with each setting's range and failure modes, is held by four companion skills a language model can apply to composition or revision: transmission, which holds the reader model and the transform, and transmission-vantage, transmission-load, and transmission-land, which hold the settings.
The three instruments and their settings are fixed by The Transmission Model Ontology, the criterion document the instruments are written against.
Baruch Fischhoff, "Hindsight is not equal to foresight: The effect of outcome knowledge on judgment under uncertainty," Journal of Experimental Psychology: Human Perception and Performance 1, no. 3 (1975). Reprinted at <https://pmc.ncbi.nlm.nih.gov/articles/PMC1743746>. On the consequences for investigations, see K. Henriksen and H. Kaplan, "Hindsight bias, outcome knowledge and adaptive learning," Quality and Safety in Health Care 12, suppl. 2 (2003), <https://pmc.ncbi.nlm.nih.gov/articles/PMC1765779>
Transmission Configurations: A Reference Roster, companion appendix, holding thirty-six worked postures with the gradients that adjust each for its nearest neighbors.