There is a widespread instinct among puzzle solvers to keep their reasoning in their heads. Writing things down feels slow, unnecessary, almost like admitting defeat. The truly skilled solver, the instinct suggests, should be able to hold it all mentally and navigate to the solution through pure cognitive power. This instinct is wrong — and acting against it is one of the most impactful improvements any intermediate solver can make.
Working Memory Is the Bottleneck
Human working memory — the cognitive system that holds active information for reasoning — can maintain roughly four to seven distinct items simultaneously. In a hard Sudoku with full candidate notation, a single box might contain thirty or more candidate digits. In a cryptic crossword, you're simultaneously holding the surface reading, the possible definition, the wordplay type, and multiple answer candidates. In a logic grid, you're tracking dozens of positive and negative relationships across multiple categories.
Trying to hold all of this in working memory while simultaneously reasoning about it exceeds what human working memory can reliably do. The result is errors — forgotten constraints, dropped candidates, contradictions that get accepted because the solver lost track of an earlier deduction. Writing things down offloads this burden from working memory to paper (or screen), freeing cognitive resources for reasoning rather than remembering.
The Audit Trail Benefit
Written reasoning creates an audit trail — a record of every step in the logical chain that led you to a current state. When something goes wrong (a contradiction appears, a placement turns out to be incorrect), the audit trail lets you identify exactly where the error occurred and backtrack cleanly to the last verified state. Without it, contradictions are catastrophic: you know something is wrong but have no way to find the source, so you may have to start over.
In Sudoku with full candidate notation, this means maintaining accurate candidate lists and updating them after every placement. When a contradiction appears, you can scan back through recent placements — each one is visible in the grid — and identify which one created the problem. In bifurcation trials, clean written notation means you can discard the hypothesis and its consequences without contaminating the verified portion of the solve.
Writing Slows You Down — And That's Good
The objection to writing things down is usually speed: it takes longer. This is true in the short run and false in the long run. A solver who writes everything down and makes no errors completes a hard puzzle faster than one who works mentally and has to restart twice. The "slow" written approach is faster in practice because errors are catastrophically expensive — the time cost of a restart or an extensive backtrack far exceeds the time cost of careful notation throughout.
Writing also forces precision. When you commit a deduction to paper, you must articulate it clearly enough to write it. Vague hunches that "seem right" become impossible to write in precise form — which reveals that they weren't actually supported by the logic available. This forced articulation catches errors before they propagate, when they're cheap to correct.
Calibrating What to Write
Not everything needs to be written. The goal is to externalise reasoning that exceeds working memory capacity and preserve decisions that will need to be revisited. For Sudoku: always maintain full candidate notation on medium difficulty and above. For crosswords: write down uncertain answers with a question mark and maintain a list of crossing letters that need to be consistent. For logic grids: always use the full grid and mark every deduction immediately. For simpler puzzles below your comfortable ability level, mental solving is fine — the written approach pays dividends precisely in the situations that most need it.