The Case Against the Crosswordese Blacklist

ERNE is a sea eagle. OLEO is margarine. ETUI is a small ornamental case for needles. Nobody has used any of these words in conversation in living memory, and all three have appeared in thousands of published crosswords, because they are short, vowel-heavy and cross beautifully.

The consensus position is that they are a flaw. Word lists ship with them scored low. Constructors apologise for them. Reviewers count them. I want to argue that the consensus is roughly right and precisely wrong, and that the difference matters.

What crosswordese actually is

The category is not really about obscurity. It is about words whose only remaining habitat is the crossword grid. ADIT, ANOA, ESNE, SNEE, ALEE - each has a real definition and essentially no real usage.

They persist for a structural reason. A 15x15 grid with 38 blocks contains a lot of three- and four-letter slots, and English has a finite number of short words with useful letter patterns. Ban enough of them and the remaining set cannot fill the grids the format requires.

You cannot ban a category of word without also banning the grids that need it.

The part the consensus gets right

A grid stuffed with crosswordese is genuinely worse to solve, for a reason that has nothing to do with taste. These entries are unguessable. A solver who has three of four letters in a real word can usually infer the fourth. A solver with three of four letters in ESNE has nothing, because the word carries no meaning they can reach for.

So each crosswordese entry converts a moment of inference into a moment of pattern-matching against memory. Do that once and it is a small tax. Do it eight times and the puzzle stops being a puzzle and becomes a vocabulary quiz on a vocabulary nobody has.

The part it gets wrong

The blacklist treats these words as intrinsically bad rather than contextually expensive, and that produces two bad outcomes.

First, it lets much worse entries through. A blacklist flags OLEO and waves past a partial like ATA or a random Roman numeral or an abbreviation that only exists in one trade journal. Those are harder on the solver than OLEO is, and no scored word list catches them, because they are not words.

Second, it makes fill quality a lookup rather than a judgement. A constructor optimising against a list will produce a grid with a good score and no character. The list cannot tell you that one obscure entry, well crossed and cleverly clued, is a pleasure, while three perfectly common entries stacked into a dead corner are not.

A better test

Replace the blacklist with three questions asked of every questionable entry:

• Is every crossing solid? An unfamiliar entry with fair crossings is a mild inconvenience. The same entry crossing another unfamiliar entry is an unfair square, and unfair squares are the actual failure.
• Can it be clued interestingly? If the only available clue is the dictionary definition, the entry contributes nothing. If it can be clued with wit, it earns its slot.
• How many are there in total? One is invisible. Two is fine. Five is a pattern, and the pattern is what solvers complain about.

The historical angle

It is worth remembering that the category is not stable. Words drift into crosswordese as usage declines and drift out as it recovers. OLEO was ordinary vocabulary in 1955. Half the tech vocabulary that felt fresh in 2005 is heading the other way now.

Which means a permanent blacklist is describing a moving target with a fixed list, and constructors who follow it uncritically will still be avoiding words that came back into use ten years ago. Dictionary usage notes are a better guide than any scored word list, and the historical grid record at XWord Info will show you exactly when a given entry started appearing and whether it is on the way in or out.

Where I land

Use fewer. Cross them cleanly. Clue them well. And stop treating the presence of a single unusual short word as evidence of laziness, because the constructors producing the most interesting grids are almost never the ones with the highest word list scores.

If you want a sense of how much short-entry vocabulary a working solver actually absorbs, the answer archives are instructive: browsing a year of daily crossword solutions or the syndicated grids at Thomas Joseph makes the recurrence patterns obvious in a way no essay can, and a good directory of solver and constructor tools is the fastest way to find the word lists worth arguing with.