Logos52
wiki / Learning Craft / Reading and Retention

Reading & Retention

technique updated 2026-08-14

Reading & Retention

Reading and retention are two separate operations, and this page is about the second. Taking material in — reading it, watching it, hearing it — puts it in front of you and does nothing else; what you keep depends on what you do afterward, and the right move differs by what kind of material it is. Procedural material wants practice. Analogous material wants you to argue with it. Conceptual material wants mapping against what you already hold, evidence wants rehearsing in the setting where you would use it, and reference material only needs somewhere reliable to sit. The target throughout is memory you can use: enough to reason with, teach from, or act on.

The balance rule

Consumption is the exposure. A book, a lecture, an audiobook, an article, a clipped source. Digestion is the afterwards: practicing, mapping, critiquing, rehearsing, connecting, or storing. Those two stages have to stay balanced. Every meaningful unit of consumption needs a matching digestion step. More input is only valuable when the system can process it. Overconsumption creates future debt. If there is no time or energy to digest, the practical move is usually to consume less.

One generic note method is the failure this rule names. Information types need different processing moves, because the encoding that matches later use is the one that wins. A pile of familiar fragments is what balanced consumption without matching digestion produces. Recognition is not usable memory.

PACER is a triage system for deciding what kind of digestion a piece of information needs. The point is effort allocation, not labeling. The five jobs, in order: practice, critique, mapping, contextual rehearsal, store-or-recall. Mixed types are the norm. An analogy can sit inside a procedure or a map. Classify the job, not the sentence. A reader who tries to label every line will stall.

Practice, then critique

Procedural information is how to do something. Digestion is practice. Coding techniques, pronunciation, a workflow, a clinical or physical procedure, a step-by-step skill rule, a study method. Procedural information decays quickly when it stays abstract. Notes can support practice. They cannot replace it. A written procedure is still declarative. If the information tells how to do something, the reading stops early enough to try the thing. Kolbs Experiential Cycle is that attempt: notice the break, name the adjustment, try again.

Analogous information is related to something already known. Digestion is critique of the analogy. The learning is in the alignment, including where it breaks. Critique is the missed step. The questions, in full:

  • What is actually similar?
  • What is different?
  • Where does the analogy break?
  • What condition makes it useful?
  • What condition makes it misleading?
  • Can it be extended, narrowed, or replaced?

Critiquing attaches the new idea to an existing network with boundaries. Comparison, discrimination, relationship-building: this is a Deep Processing move applied to a resemblance.

Knowledge is not a line

A page presents ideas in a sequence because language is sequential, but the knowledge itself is not a line. Expert knowledge is flexible. One concept can lead to another through multiple paths. That is why conceptual digestion is mapping rather than more notes in the author’s order, and why the remaining three jobs allocate effort instead of adding labels. The letters are a way of spending attention. They are not a finding about how memory is built.

Map, attach, store

Conceptual information is principles, explanations, relationships, theories. Digestion is mapping. Good conceptual digestion groups, names relationships, separates central from supporting, moves ideas as understanding changes, preserves dependencies and contrasts, and builds a reconstructable structure. Mapping means rebuild the relationships. It does not mean a particular format, and it does not mean buying a tablet.

That work, while the source is still live, is the Bear Hunter System: four-pass encoding. Aim writes what the material must answer before it is opened. Shoot works the source against those questions into a rough map. Skin prunes the map to what can be reconstructed without the source.

Evidence is examples, cases, dates, statistics, studies, named incidents. Digestion is store and rehearse in context. Evidence should be captured reliably. It should not dominate the reading session. Its value comes from what it supports. Evidence is not trivia. It is proof material attached to a conceptual structure.

Good evidence rehearsal asks:

  • What concept does this support?
  • When would I use this example?
  • What claim does it make concrete?
  • What comparison does it enable?
  • What question could this help answer?

Those questions are the rehearsal form. Cards of dates are the wrong later test. Evidence stays on the map unless the later test is the fact itself. Spaced Interleaved Retrieval handles reference, and handles evidence only when the evidence itself must be produced.

Reference is exact details, labels, names, values, lookup facts. Digestion is store and rehearse directly if memory is required. Many facts only need a clean store. Premium reading attention is not spent rereading exact details while conceptual and procedural work is unfinished. Reference is usually easier to handle later, on the same spaced loop.

Research Foundations

The working loop is consume, classify, matching digestion, encode into a usable structure, rehearse or apply later, retrieve in the form the work requires. The load-bearing step is the stop: when digestion falls behind, input slows. The better question is not how much can be gotten through. It is: what kind of information is this, and what does it need before the next unit?

The failure is familiar. Consume too much, fail to classify, apply one generic note method, forget or settle for shallow familiarity, reread later, pay the same cost with interest. Rereading is a low-utility strategy. Familiarity mistaken for mastery is the metacognitive error. The visible session looks full. The durable gain stays thin. The repair: slow the input, identify the job, digest before continuing, store only what should be stored, rehearse only what needs rehearsal.

Good reading feels selective and controlled. It slows at important points. The next action is obvious. Concepts form a map. Evidence attaches to claims. Procedures become attempts. Details are stored without hijacking the session. Less is consumed, and more survives.

Warning signs: the source outruns the map; everything is a note; details steal attention from structure; procedures are preserved but not tried; analogies are noticed but not tested; familiarity without usable memory.

Inside this vault, Raw to Wiki Compilation should treat PACER as a source-processing lens. The Bear Hunter System should handle procedural, analogous, and conceptual material while the source is still active. Spaced Interleaved Retrieval should handle reference, and evidence only when the evidence itself must be produced. Knowledge Base as Thinking Partner helps classify and digest. It must not replace first-brain digestion. Are You Thinking, or Just Consuming? is the warning light. A knowledge base collecting sources without matching digestion is elegant overconsumption.

What is established: what is done with the input decides the trace; the encoding that matches later use wins; procedural knowledge is built by doing; analogical learning is in the alignment, including the break; expert knowledge is a network; rereading is low-utility and testing is high; spaced recall is the tool for arbitrary items. What is not established: PACER as a field taxonomy. The five moves are supported. The five-letter partition is house triage. Mapping-must-be-a-mind-map is mixed; the instruction is relations, not a format. Whether classifying during the read outperforms a second-pass sort is open.

The case against: labeling every sentence stalls the read; using the five letters as boxes instead of jobs; letting evidence become a card pile; treating mapping as a format purchase. The price is digestion time — that is the whole point — and the benefit in the same breath is a usable structure instead of familiar fragments. Quit signals: the only output is notes and there is no next action; procedures still untried after a full source; two sessions of “I’ll digest later.” Checkable: less consumed, more survives; a procedure was attempted; an analogy was tested at the break-point; evidence is attached to a claim; a map can be reconstructed with the source closed.

Input slowed so the afterwards could finish. What remains is usable. That is the same split the page opened on, seen from the far side of a shorter stack.

Open Questions

Whether classifying during the read outperforms a second-pass sort.

Sources

  • Craik & Lockhart, “Levels of processing,” Journal of Verbal Learning and Verbal Behavior, 1972; Craik, 2002.
  • Morris, Bransford & Franks, “Levels of processing versus transfer appropriate processing,” Journal of Verbal Learning and Verbal Behavior, 1977.
  • Anderson, ACT-R; Fitts & Posner, Human Performance, 1967.
  • Gentner, “Structure-mapping,” Cognitive Science, 1983.
  • Chi, Feltovich & Glaser, “Categorization and representation of physics problems by experts and novices,” Cognitive Science, 1981.
  • Fiorella & Mayer, Learning as a Generative Activity, 2015.
  • Dunlosky, Rawson, Marsh, Nathan & Willingham, “Improving Students’ Learning With Effective Learning Techniques,” Psychological Science in the Public Interest, 2013; Roediger & Karpicke, 2006.
  • Cepeda, Pashler, Vul, Wixted & Rohrer, “Distributed practice in verbal recall tasks,” Psychological Bulletin, 2006.