Mindmaps

The first marks on a mindmap are three to seven guessed chunks, placed far apart so the relations show and the next layer has room, then redrawn once a better grouping appears. Headings copied off a table of contents and given branches stay notes in a new shape. The work is those decisions.

Non-linear note-making keeps the work of deciding how pieces relate, which a verbatim copy of the speaker’s order throws away. The form is a wait until a relation can be named, then a mark that says how. A first page that still follows the contents list is a form error rather than a verdict on the method. In a live session the same decision runs at the speaker’s pace: only the sentence that restates each finished point gets written.

Core takeaways

• The first marks are three to seven guessed chunks, placed far apart, then redrawn when a better grouping appears. The backbone they form is the most important thing on the page.
• Wait until a relation can be named in one sentence or one arrow before anything goes on the page. If the sentence will not come, the next move is another pass at the material, not a page of marks.
• Default to arrows. Choosing between one-way, reverse, double-headed, a bracket, and a deliberate plain line is the processing; a plain line only says that a connection exists.
• Spiderwebbing, waterfalling and wheel-and-spokes are one condition, no lateral relationships, and the usual origin is chunking in the material’s own shape. The repair is to flatten the headings to a list and let the chunks fall out of the relations that are there.
• In a live session, listen until the speaker finishes a point, then write the shortest restatement that can be managed. More words copied means fewer decisions about what mattered.
• The reachable record supports learner-constructed maps with labelled, directed links against conventional study, with a larger effect for building a map than for studying one somebody else built. Verbatim capture is what fails, not the keyboard.

Wait for the relation, then mark it

The strongest habit is the delay. After a section, the relation gets said in one sentence or one arrow before any notes exist. If that sentence will not come, the next move is another pass at the material, not a page of marks; the map is not the bottleneck. These mechanics serve Shoot, the encoding step that works a source against questions into a rough map.

Continuous note-taking feels safe. It also adds a later revision cost: the deciding was skipped, so the same material has to be processed again, and the total time goes up. The full page that felt secure costs that revision time after the session.

Sitting with the material before a mark hits the page is uncomfortable. That discomfort is the work happening, not a score of how well it is happening. Writing to end the feeling chooses the cheaper-feeling path that defers the work.

The default mark is an arrow, not a line; choosing which kind of arrow is the higher-order work a bare line lets a page skip. Concepts go spatial; specific facts often stay verbal. Forcing every fact into a drawing is not this technique. A mark on the page reinforces a relation already decided. It is not a substitute for grouping, which happens first.

Chunks form first, then the map. Groups come from relations and importance, not from the material’s hierarchy. The way a source hands over its headings already decides the first groups, and those given groups already look logical, which is why “there is no other way” feels true. The headings are a starting list, not the chunks.

A page that lasted is a page that decided a relation before the mark. After one delayed pass the arrows can be defended: what kind of relation, which way. If the page is neater and still in the material’s sequence, the delay did not happen.

In a live session

A live session is used by arriving holding the topic’s few largest ideas and the flow between them, then writing only the sentence that restates each finished point. The session then reinforces understanding that already exists instead of demanding it with nothing prepared. Prestudy is the setup pass itself, how the topic-level pass before the session is run. Study Scheduling is where that block goes on the calendar, so the pass exists before the session rather than after it.

The work is the compression. Verbatim capture is what fails, not the keyboard.

What helps is whatever forces a choice about what to keep. Listen first. A point that has not been finished cannot be paraphrased, so writing while the speaker talks can only be transcription. The paraphrase has a spec: the simplest, shortest form that can be managed.

The speaker finishes a point. The pen does not move. The point gets said back in one sentence. That sentence is what gets written.

Word count is a proxy. More words copied means fewer decisions about what mattered. Learning happens in that judgment, not on the page.

A relationship that costs a sentence in prose costs one arrow on a page, and where a thing sits becomes its own cue. The note format assumed in the room is relationships first, not the speaker’s sentence order.

The effort of getting a point into shorter words of one’s own is the signal. Anxiety that produces more copying is not. Struggling to process before writing points to the processing, not a trait, and is the thing the delay repairs. Delay the pen as long as the point can still be held, then extend the window. Relating each point to the big picture and grouping as it arrives is what makes the delay possible. If delaying the pen produces panic and more copying rather than a sentence of one’s own, shorten the window instead of extending it.

These are cheap habits that raise the least a session yields. They do not replace studying the material afterwards, and they are not the strongest note-making methods available. What is being trained is the processing, with the room as the place it gets tested under time pressure. The checks are self-reports.

Building the map

The first marks are three to seven big chunks, treated as guesses rather than commitments. That range is this system’s starting heuristic, not a finding about how much a head can hold. The chunks get placed so the relations and the flow between them are visible at a glance. That backbone is the most important thing on the page.

Sub-concepts come next, each one tied back to the big picture as it lands, in the same sequence Layers of Learning uses on any other page: backbone, then concepts, then important details last. A highly specific fact with no relational hook does not belong on the map. It belongs in Flashcards, and it is tested later through Spaced Interleaved Retrieval.

Space is planned, not discovered. The first chunks sit far apart so the next layer has room to grow. A map that cannot be cleaned usually started with those first chunks packed close, so everything after had to be threaded between them. A long link that weaves between other nodes is hard to remember as a relation. The repair is to move the point, not to route the line: related nodes sit physically near each other so the link is short. Proximity is a form of the relation, not a layout preference.

The backbone stays visually dominant. Everything on the page stays provisional. A first map is expected to come out tangled. The second pass is where the spatial work happens, and that pass is planned for rather than tolerated. Rearranging is not a loss of the first map’s work. It regularly surfaces a chunk that did not exist in the first version, a pattern invisible while the content was still being placed.

A copied map is readable on sight. Arrows added after the content was written run through words, take long detours around nodes that were placed first, and cross each other. A map whose relationships were decided first has short, direct links, because the nodes were positioned to make them short. The failure is near-universal for a simple reason: content is easy to place, so it gets placed first, and the relationships are then fitted into whatever space is left.

Importance, then arrows

Grouping by surface similarity is not enough. The similarity has to matter enough to justify the group. Novices group by surface features; experts group by principle. Finding the right chunk often means finding the root reason something matters, not the immediate one. Ask “so what?” until the question no longer makes sense; that answer is usually a better chunk name. Genetic diversity matters because it produces variation: that is the immediate reason. What variation buys is survival. Survival is the better name: broader, able to hold more under it, easier to form relations to and from. The full craft of that move lives on Importance-Based Chunking.

The first chunk structure is almost never the simplest it can be. A missing perspective, or a simpler way, is assumed to be there. Finding it is difficult. Find it anyway. Groups do not have to be the material’s groups, or anyone’s official ones. A grouping earns its place by making the topic easier to hold.

Arrows force a decision about what kind of relationship is there. Lines only acknowledge that a connection exists.

Five nodes joined only by plain lines.

Every directed arrangement over those same five, one way, the other way,
both ways, a bracket that groups rather than points, is consistent with
that picture. The lines are true, and they hide every distinction that mattered.

The vocabulary in use is one-way, the reverse, double-headed, a bracket for grouping, and the deliberate plain line. Choosing among them is the processing. The choice cannot be made without deciding what kind of relationship is there. A plain line is the right call at the lower levels of detail, where a chunk splits into component facts that genuinely co-exist without direction. Default to arrows. Defaulting to lines is the lower-processing option, and it should be a deliberate choice rather than a habit.

The failure family: spiderwebbing, waterfalling, wheel-and-spokes

The three named failures are one condition, no lateral relationships, seen from three angles.

Doing it right is checkable on the page: new relations get easier to see as the map grows, each piece still connects to the big picture, and the first version is messy and then organizes. The common failures are a missing backbone; more than about four branches from a single node, which this system treats as a sign that further chunking is available rather than as a working-memory finding; chunking by surface similarity; lines where arrows were needed; treating the first structure as final; and the three forms below.

Spiderwebbing is chunks formed first, usually taken straight off a source’s headings, with relationships added afterwards. The layout was never built to hold those relationships, so the arrows scatter, run long detours, and cross. The name comes from how the links look strung between branches. The cost lands four ways: the map takes longer to make, holds less, loads working memory while the layout is being fought, and is frustrating enough that mapping gets abandoned as a technique that “does not suit” the person doing it.

Waterfalling is the same condition seen from the other side. All connections flow downward in a fixed hierarchy, with no lateral relationships. When laterals are then forced in, the arrows cross and the map becomes unworkable. The second form of the same condition is everything funneling through one central node. The consequences are specific. Maps of unrelated topics come out looking identical to each other. Every terminal fact has to be memorised, because nothing in the structure holds it.

Wheel-and-spokes is that funnel with one hub. Chunking can be good and the map still fails, because every arm hangs off the centre and none of the arms touch.

The repair, in order: flatten the material’s headings into a plain list of keywords. The flattening is what breaks the material’s framing. Find the relationships and similarities across the list. Judge which of those similarities matter enough to group by. Let the chunks fall out of that judgement. Chunks appear at this stage that were in no heading of the material. If the arrows on a finished page all run the same direction as the chapter headings, the grouping never left the material: flatten to a list and start again.

A map already built does not have to be thrown away. Find the chunk carrying the highest density of connections and re-chunk it. Grouping the concepts that all connect to that node removes the need for most of the lines, and it preserves every relationship already found. The four standing checks against the family: whether the current structure is the simplest and most intuitive one available; whether chaotic links can be brought closer or into the same chunk; whether a structure has been committed to too early; whether each link is necessary, and what kind of relationship it is expressing.

One common miss is justifying a node by what it produces rather than by what gave rise to it. That miss is owned by Reverse Causality, and it shows up as a signal of its own: calling A important only because it produces B, so both now have to be memorised. The repair is to anchor importance in a wider network, so each piece is inferable rather than rote. Intuitive chunking tends to prevent the loop; importance-checklisting and simple Q&A can worsen it. Importance Based Chunking is what makes a chunk a chunk; Survive and Thrive is why relations are the keep-signal.

Chunking in the material’s own shape is the most common origin of the whole family, because the shape arrives already looking logical. The delay, waiting until a relation can be named before any mark is made, is what keeps that shape from being written down as if it were the chunks.

The finishing check: GRINDE

GRINDE is this system’s six-point check, applied during the Skin step of Bear Hunter System, that a finished map is grouped, in the learner’s order, laterally linked, spare with words, flowing, and visually ranked. It is a house checklist defined by those six qualities, not a published construct.

Letter	What it checks
Grouped	Chunked by importance. Backbone clear. Layers in order. A node carrying more than about four branches is a node with more chunking still available in it: a sign, not a law.
Reflective	The order of relations, logic, concepts, and details is how the topic is understood, not the order it was presented. Layers run backbone, then concepts, then important details. A map that reproduces the material’s numbering comes out looking like a family tree: chunked, tidy, and still the lecture.
Interconnected	Arrows connect within layers as well as between them, and only where a real relationship exists. Arrows added because two words sit near each other are not connections. A map can fail this check by having too many links as easily as too few.
Non-verbal	Fewer words leave the structure visible instead of buried under text. A mark that stands for the idea is the target. Drawing everything is not.
Directional	A clear flow of how concepts influence other concepts. Direction can only be drawn once a decision has been made about which of two things outranks the other.
Emphasised	Visual hierarchy by size, colour, or weight. The emphasis forces an explicit ruling on which relationships matter most, and it lets the backbone be picked out at a glance on every later pass.

The mark under Non-verbal has to distinguish, not decorate. A set of drawings that all look alike is worth no more than the words were. A symbol earns its place when it carries the attribute: a shape drawn to say strong, a small sun inside a letter to say this phase is the relaxed one.

A map built in the right order comes out visibly lopsided: the region drawn first is neater and denser, later additions trail off in one direction. That skew is a side effect of building from the backbone outward, and it is a good sign rather than a defect. The finishing audit runs letter by letter and comes out with named improvements: this cluster needs another level of grouping, this region needs a mark instead of a word, this main line needs to be heavier. A finished map is a map whose remaining faults have been named.

Links do not cross. Where two must, one of the points is in the wrong place: pull it out rather than routing the line around the obstruction. Take out what has become obvious. Facts written down early are, by the end of the topic, things that will not be lost, and they can leave the map. Other facts leave because they never connected to anything. Those are two different exits. Finishing does not remove information; it changes what can be seen. Overlapping chunks, when spotted, add a dimension rather than a contradiction: the same material admits two valid groupings at once. No copy-pasted diagrams. No lined or cramped paper, if the surface can be chosen.

Keeping citations without losing the map

Non-linear encoding makes citation awkward, so two layers run in parallel. The map is for understanding. Every two or three sources, not after every page and not at the end of the topic, a short linear, cited synthesis is for consolidation and later retrieval of sources. The cited-as-you-go synthesis is the method for a single piece of writing that has to ship. A stored reference layer, in a reference app or a citation manager, is the method for sources reused across projects. When both jobs exist, the two combine; neither replaces the other.

Both methods add administrative time on top of encoding. The delay is also slower in the first ten minutes. That is the price, and the benefit sits next to it: the map holds relations instead of a transcript, and the citations still exist when a paper is due. The stored layer is worth a trial on a handful of sources before a whole field is committed to a template; a bad structure is expensive to unwind. Being able to cite the right source later is not a filing problem. It rides on already having a relational grasp of the topic, plus using those sources often enough that retrieval is interleaved with the writing. A well-tagged library on a thin map still will not hand over the right paper.

Where the linear form is still right

Verbatim linear notes are the wrong tool for understanding, not for a timestamped record. Minutes, a quote log, a legal pad: those jobs keep the linear form. A first map that copies the table of contents is not this technique either, and it is the usual reason a first try gets abandoned.

What the research supports

What the reachable record supports is learner-constructed concept mapping against conventional study, with the effect larger for construction than for studying a map somebody else built. That finding is the claim that the deciding is the encoding, measured from the outside. The radial-diagram brand of mind mapping is the weaker literature: small, mixed studies. The evidence sits with maps built by the learner with labelled, directed relationships, not with the radial format as a format. Dual coding supports a mark alongside words. It does not support a blanket claim that images encode more than text.

Verbatim capture is the weak form. Typing is a risk factor for that drift, not a finding that the keyboard or the line is the enemy. Constructing a relation structure, nodes plus labelled links, is what the outside record actually supports.

Finding a better chunk structure is slow and heavy. Re-chunking a topic is a real second pass. It loads working memory while the old structure and the candidate one are held together, and the first several attempts at a topic feel wrong before the move transfers. What it buys, at the same time: a structure that expresses the same relationships with far fewer links, and one unique enough to a topic to be memorable.

A finished map is not evidence that anything was learned. A map made without the deciding is worth no more than any other notes, and it costs more to make, which is exactly the experience behind “mind maps don’t work for me.” A second honest failure: a map can prioritise relationships so hard that the detail never arrives, leaving a clean shape with too little in it.

Two quit signals, each with a next move. If the nodes on a finished map cannot be justified, the drawing is not the problem: stop mapping and go back to deciding importance. If a large share of a topic is lost when it is tested from memory, suspect the structure rather than memory: check the map for the funnel and the scatter, and re-chunk the densest node.

Three things should be checkable on the page itself: new relationships get easier to see as the map grows, the map comes out skewed, and what is left to memorise outright is a small set of specific facts rather than the topic. How wide to build that structure is the job of Best-attempt Encoding.

The deciding is the encoding. A finished page says which order the work ran in. Short, direct links, a visible backbone, a lopsided first region, leftovers off in flashcards: the map was built by judging. Arrows that run through words, detour, and cross: the headings were placed first. The simpler perspective, once found, is visible the same way.

How to practice this

1. Start with three to seven guessed chunks, placed far apart, and treat them as guesses. Notice whether the next layer has room and whether each link stays short.
2. After each section, say the relation in one sentence or one arrow before any notes exist. If the sentence will not come, go back to the material rather than to the page. Notice that the discomfort of waiting is the work happening.
3. Default to arrows and name each one: one-way, reverse, double-headed, a bracket, or a deliberate plain line. Notice which choices cannot be made without deciding what kind of relationship is there.
4. In a live session, wait until the speaker finishes a point, then write its shortest restatement. Notice the word count: more words copied means fewer decisions about what mattered. If waiting produces panic and more copying, shorten the window.
5. On a finished map, find the node carrying the most connections and re-chunk it. Notice that most of its lines disappear while every relationship already found stays.
6. Run GRINDE letter by letter and write down the named improvements. Notice that the map comes out lopsided, with the first region neater and denser; that is a good sign.

Related pages

• Bear Hunter System: the encoding loop whose finishing pass this checklist serves.
• Deep Processing: the parent dimension, encoding that changes the structure, not the volume.
• Importance-Based Chunking: what “grouped by importance” actually is and what makes a chunk a chunk; one instance is kept here and the full craft lives there.
• Spaced Interleaved Retrieval: where leftover unconnected facts go, and how the finished map is tested.
• Best-attempt Encoding: how wide to build the schema.

3 more related pages

• Shoot: the encoding step the delay and the mark serve.
• Schema: why relations are the keep-signal.
• Reverse Causality: the A→B loop that adds a memory burden.

Open questions

• How far a typed map can carry the same deciding, now that paper is no longer the default surface.

Sources

• Nesbit, J. C., & Adesope, O. O. (2006). Learning with concept and knowledge maps: A meta-analysis. Review of Educational Research, 76(3), 413–448. 55 studies. Learner-constructed concept mapping against conventional study, d ≈ 0.46; the effect is larger when learners construct rather than study a provided map. Construction of a relation structure versus conventional study; supports the operating claim from the outside.
• Chi, M. T. H., Feltovich, P. J., & Glaser, R. (1981). Categorization and representation of physics problems by experts and novices. Cognitive Science, 5(2), 121–152. Novices group by surface features, experts by principle: the supported basis for chunking by importance rather than by resemblance.
• Farrand, P., Hussain, F., & Hennessy, E. (2002). The efficacy of the ‘mind map’ study technique. Medical Education, 36(5), 426–431. Small and mixed. The weaker literature around the radial-diagram brand, cited as the bound rather than as support.
• Paivio, A. (1991). Dual coding theory: Retrospect and current status. Canadian Journal of Psychology, 45(3), 255–287. A mark alongside words has a trace of its own. The theory does not license “images encode more than text.”
• Mueller, P. A., & Oppenheimer, D. M. (2014). The pen is mightier than the keyboard: Advantages of longhand over laptop note taking. Psychological Science, 25(6), 1159–1168. Mechanism: laptop notes were more verbatim; conceptual questions suffered. Notes that drift toward verbatim retain less. The mechanism survives; the tool claim does not, and the format finding did not hold in later replications.
• Morehead, K., Dunlosky, J., & Rawson, K. A. (2019). How much mightier is the pen than the keyboard for note-taking? A replication and extension. Educational Psychology Review, 31, 753–780. Failed to reproduce the conceptual longhand advantage; the longhand-beats-laptop tool claim failed to reproduce.
• Urry, H. L., et al. (2021). Don’t ditch the laptop just yet: A direct replication of Mueller and Oppenheimer (2014). Psychological Science, 32(8), 1192–1206. Direct replication; no significant longhand advantage.
• Ausubel, D. P. (1960). The use of advance organizers in the learning and retention of meaningful verbal material. Journal of Educational Psychology, 51(5), 267–272. Arriving with a more abstract frame makes later detail land. Mayer’s pre-training principle is the later statement of the same move.
• Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11(6), 671–684. Encoding quality over repeated exposure.
• Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students’ learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4–58. Elaborative interrogation and self-explanation.
• Miller, G. A. (1956). The magical number seven, plus or minus two. Psychological Review, 63(2), 81–97. Grouping as load reduction. Cowan (2001) is the later working-memory bound.