↩ ExploRational · A Catalogue of Divergent Rationality
explorational · a divergent move
What is opaque in prose can be obvious in a picture, and what hides in a picture can leap out of a table. If the representation does reasoning the notation cannot, which body should your problem wear next?
In the winter of 1953 the structure of DNA was hiding inside a pile of X-ray photographs and a wall of chemical equations. Watson and Crick did not solve it by staring harder at the algebra. They cut the bases out of stiff cardboard and bent metal plates into a scaffold, and they built the molecule with their hands — and the model did arithmetic the equations had only mumbled. A shape that balanced; bonds that closed one way and not another; a helix that had to be double because the cardboard would not lie flat otherwise. The tin and cardboard were not a picture of the answer. They were the instrument that found it.
Every medium computes something for free. Prose tracks who wants what and who stands in the way; a diagram exposes what connects to what; a table forces you to name the two axes a mess varies along; a timeline settles what came before what. Hand one stuck problem to all four and each returns a different free inference — and the thing that will not yield in one notation often falls open in another. The bench below is four blank bodies waiting for your problem to try each of them on.
the signature instrument
State your problem in one line, then wear it four ways: type it as prose, draw it as boxes and arrows, force it into a table's rows and columns, lay it along a timeline. After each, log the one thing that became visible. The readouts are structural and honest — canvases visited, entries logged, real time spent on each. The bench counts and times; it never grades. Load the worked example to see one scheduling dispute run through all four.
Tell it as a story. What wants something here? What blocks it? Who acts — and what happens next? Prose is the medium that tracks desire and obstacle.
Draw it. Shapes and arrows only — boxes for the parts, lines for what connects to what. A diagram indexes by location: it puts the relationships where your eye can read them off.
This canvas is not interpreted. No shape is read, scored, or judged right or wrong — the bench only records that you drew. What you see is exactly what you drew.
What are the rows? What are the columns? Naming the two axes is the instrument — the grid is empty until you decide what the problem varies along. The forced question does the work; the cells are only the afterthought.
What happened before what? Put the events in order. The strip forces the sequence prose can blur with a soft "meanwhile" — and causation often hides in exactly the order you left vague.
The bench times you; it does not grade you. Where you lingered is data about where the problem lives — a canvas you could not leave is often the canvas where the difficulty actually is.
the medium computes
A representation is not neutral packaging. Each medium has affordances — operations it makes cheap and operations it makes impossible — and those affordances do part of your reasoning before you have consciously begun. Larkin and Simon put it precisely in 1987: a diagram can be worth ten thousand words not because it holds more information but because it indexes that information by location, so that inferences a sentence would have to spell out are simply read off the page. The medium is a silent collaborator with opinions.
The table's opinion is a question you cannot dodge: what are the rows, and what are the columns? A grid stays empty until you name two dimensions, and naming them is the move — the instrument is the forced choice, not the cells you fill afterward. A surprising share of stuck problems come unstuck the moment you are made to say aloud the two axes they vary along. The timeline is just as coercive about a different thing: order. Prose blurs sequence with a soft "meanwhile" or "around that time"; a timeline strip refuses. It demands you commit to what happened before what — and causation frequently hides in exactly the sequence you had left vague.
what to try
Run a genuinely stuck problem through all four canvases without deciding in advance which will help. Watch which medium's forced question you resist answering — the resistance is itself a clue about where the problem is soft.
Read your own noticing log as the product. Four short lines, each naming what one body of the problem made visible, are worth more than any single pretty diagram. The log, not the artifacts, is what you carry away.
Check where the dwell timer says you lingered. The bench times you and refuses to grade you — but the canvas you could not leave often exposes the problem; the number points to it.
tin and cardboard
Watson's own account in The Double Helix is candid about how physical the thinking was. The pair kept rebuilding the model because the model, not their intuition, was doing the checking: a proposed pairing either let the pieces close into a stable shape or it did not, and the cardboard was a harsher referee than any argument. This is representation as externalized reasoning — offloading part of the cognition onto a medium whose own rules grind against the problem and reject configurations the mind alone would have entertained for weeks. The equations held every one of the same facts. What the equations lacked was a body that would refuse to fit.
That is the general lesson beneath the specific triumph. When a problem will not yield, the bottleneck is often not missing information but the wrong body for the information you already have. Change the body, and reasoning you could not perform in the old notation becomes something you can nearly do with your hands. The move opens perception by shifting the medium — and perception, not more data, is frequently the thing that was scarce.
the honest bench
This bench is deliberately modest about what it knows. It measures three things and every one is structural: how many of the four canvases you have visited, how many noticing-log entries you have written, and how long you dwelt on each canvas. It counts; it never grades. It does not read your sketch — no shape is scored, no arrow judged right or wrong — and it says so on the canvas itself. It does not rate your prose or approve your table. A tool that pretended to know whether your diagram was good would be lying, because the whole premise is that the value lives in what became visible to you, which no counter can see.
So the deliverable is not a score. It is the noticing log: four honest lines about what each body of the problem revealed. The dwell timer is offered in the same spirit — not as a productivity metric but as evidence about where your attention actually went, which is often where the difficulty actually is. Read the numbers as instruments pointed at the problem, never as marks on your work.
the move ↔ the bench
| The move | The bench |
|---|---|
| the problem | the thing that will not yield in the notation it currently wears — the reason you are at the bench at all. |
| a canvas | one medium loaded with its own free computations: prose tracks want and obstacle, the grid tracks axes, the strip tracks order. |
| the forced question | what the medium makes you decide before it will help — the rows and columns, the sequence, the shapes. The demand is the instrument. |
| the noticing log | what each body of the problem made visible that the others hid — the actual output of the bench. |
| dwell time | where your attention would not leave, which is usually a map of where the problem actually lives. |
| the side-by-side | the same problem seen in the round — four bodies at once, so no single medium's blind spot goes unchallenged. |
how this opening fails
risk · novelty for its own sake
The move has a failure mode built into its pleasure: redrawing feels like progress, and there is always one more medium to try. At some point the question is not "what would this look like as a Sankey diagram?" but "what am I now going to do?" Re-representation earns its keep only when it ends in a commitment the old notation was blocking. If you have run four canvases and the fifth is a way to avoid deciding, you are procrastinating in a lab coat — a fresh medium instead of a decision.
risk · the terminating answer
The noticing log has a valid entry that closes the loop: nothing new became visible here. Some problems really are already in their best body, and a medium that adds nothing is data, not failure. Recording the null result is what keeps the bench honest — it lets you stop, rather than forcing a fake insight to justify the exercise. A log of four "nothing new" lines is a complete, respectable answer: this problem does not need re-representation, and now you know.
risk · the prettiest is not the truest
A clean diagram persuades out of all proportion to its accuracy. The medium that produced the most beautiful artifact is not thereby the one that computed the most truth; sometimes the ugly table is right and the elegant sketch is a flattering simplification. Beauty is an affordance too, and it affords overconfidence. Trust the representation that survived its own forced question — the table whose axes you could actually name, the timeline whose order you could actually defend — not the one that merely looks best on the page.
When the problem won't yield, change its body.
can you use it?
RECOGNITION — Which situation calls for the move? A: A dispute you have written and rewritten in prose for a week stays stuck. B: Your diagram is finished but ugly, so you redraw it cleaner. C: The analysis is missing a number you could get by asking.
A. The move is for a problem that will not yield in the body it wears. C is missing information, and no new body supplies data; B swaps media for beauty — the prettiest representation persuades out of proportion to its accuracy.
THE NEAREST NEIGHBOR — The table’s forced question — name the rows and columns — is the sibling move Finding the dimensions. What distinguishes this move from it?
This move changes the medium itself, and each medium brings its own forced question: axes for the grid, order for the strip, connection for the sketch, want-and-obstacle for prose. Axis-naming is one canvas of four.
PRODUCTION — Two departments claim the same budget line; the email thread is forty messages deep. Pick a canvas prose is failing at and name its forced question, before looking.
Timeline: which commitment was made before which — the order the thread blurs with “meanwhile.” Or a table: rows the disputed expenses, columns whose mandate covers each. Yours works if you named the medium’s forced question and can log what became visible — “nothing new” included.