Pencil discipline: why your wrong answer is the blockage

A corner that will not move is almost never a corner of impossible clues. It is a corner containing one wrong answer that you are still treating as settled. The clues around it are behaving exactly as they should, refusing to accept a letter that does not belong there, and the longer you stare at them the more certain you become that the setter has been unfair. The blockage is yours. This guide is about finding it quickly and removing it without dismantling the rest of the grid.

Why a wrong answer sticks

Wrong entries survive for two reasons. The first is that they are plausible. Nobody writes nonsense into a grid. The word you put in matched the clue closely enough to feel solved, which means it has a defence ready every time you look at it.

The second reason is that you stop reading its clue. Once an answer is written, the clue above it becomes a caption rather than a question. You glance at it, recognise it as the clue you already solved, and move on. That is why rereading a stalled corner from the top rarely helps. You are not rereading the clues. You are rereading your memory of having solved them.

Both effects are strongest for answers written in early, from a clue you liked, with no crossing letters to check them. Speed at the start of a puzzle is bought with risk at the end of it.

The symptoms of a bad letter

You can usually tell a wrong answer from the behaviour of the squares around it rather than from the answer itself. Four signs are reliable.

• An impossible pattern. A slot reading _ Q _ T or _ _ V T is not a hard word. It is a word that does not exist, and one of the crossing entries is wrong.
• A corner where everything needs a proper noun. If three clues in one region all seem to want obscure names, the region is being distorted by a letter that forbids the ordinary answers.
• An answer that fits the definition but not the part of speech. A clue in the past tense with a present-tense answer is wrong even when the meaning is right.
• A phrase that would need a spelling you would not defend. If you are quietly hoping the setter has allowed a variant spelling, they have not.

The same diagnostic works in reverse when the puzzle is a cryptic. If an answer satisfies the definition but you never found the wordplay, it is a candidate, not a result. A cryptic clue defines the answer twice, and one confirmation out of two is a coin toss.

The drill for finding it

Work the suspects in order of how they entered the grid, not in order of how hard the clue looked.

Step five sounds wasteful and is not. Removing one letter leaves the rest of the wrong answer in place, and the remaining letters keep steering you back to it. Clearing a block of six or eight squares costs a minute and returns the corner to the state it had before the error, which is the only state in which the clues work properly again.

What earns a letter in the first place

Pencil discipline is mostly a rule about when an answer is allowed into the grid. A reasonable standard for a solver who wants to stop losing corners looks like this. Write an answer freely if it is checked by two letters you already trust. Write it if it is a fill-in-the-blank, a proper noun or a number that admits no alternative. Write it if the enumeration, the grammar and the definition all agree, which is why the habit of reading the enumeration before the clue pays for itself twice.

Everything else goes in the margin. Write the candidate beside the grid, note which slot you meant it for, and wait for a crossing to confirm it. This is not timidity. It is the same reasoning behind working outward from checked squares in solve the crossings first, applied to the one decision that actually costs you the puzzle.

Solvers who use an app have a version of this problem that paper solvers do not. The check button removes the diagnosis along with the error, so you learn nothing about which kind of answer you over-trust. If you use it, use it at the end of the session rather than during it, and look at what was wrong rather than only at what turned red. The tools and the ethics around them are discussed in paper or app.

Building the habit

Two exercises fix this faster than general practice. The first is solving a weekday puzzle in pen, with a rule that nothing goes in unchecked. It is slower for about a week and then it is not slower at all, because you stop paying the recovery cost. The second is reviewing a completed grid against the published solution and marking not the wrong answers but the moment each one entered: unchecked, half checked or fully checked. Almost every error will be in the first category.

A syndicated daily is the right size for this because the grids repeat their habits, and the solutions are easy to check afterwards. Weekday New York Times puzzles work well, as does a Wall Street Journal grid checked afterwards against a published WSJ answer list. For a sense of how often a single letter decides a whole region, the constructor and editor commentary at the Wordplay column is instructive, and the entry data behind American grids is collected at XWord Info. The vocabulary for describing all of this, from checked squares to crosswordese, is in the glossary.