The Setter's Art
Every crossword setter faces one problem: how do you guarantee the solver a way in? Get it wrong and the grid is either trivial or impossible.
In the late 1920s two people answered that question, and answered it in opposite directions. Everything that separates a British crossword from an American one follows from the fork.
The two answers
In America, Margaret Farrar's rulebook required that every letter be checked - every white square belonging to both an Across and a Down. Miss a clue entirely and the crossings will still hand you the answer, letter by letter. Certainty, guaranteed structurally.
In Britain, Torquemada went the other way. Around 1928 he abandoned black squares for bars - and, contrary to how it's usually told, he did it to increase checking within his own tradition while freeing his vocabulary from the tyranny of the interlock. British grids settled at roughly half the letters checked. The solver is given less help, and in exchange the setter is released from having to fill every corner with whatever fits.
The trade is exact, and it's the whole story:
Britain bought vocabulary quality with unchecked letters. America bought certainty with crosswordese.
Ximenes diagnosed the American side of that bargain in 1966 as "the spate of emus and gnus, of ores and odes, of ohms and ergs." The New York Times' current submission guidelines are still fighting the same list - ERNE, ASTA, ARETE, YSER. Full interlock demands filler words, and ninety years of editors have been trying to stamp out the words their own rules require.
Torquemada
Edward Powys Mathers took his pen name from the first Grand Inquisitor, which tells you how he saw the relationship with his solvers.
He made his first fully cryptic puzzles for friends; one was submitted without his permission to the Saturday Westminster Gazette, which asked for more. When that paper folded he moved to the Observer, where his first puzzle - "Feelers" - appeared on 14 March 1926. He produced 670 of them before his death in 1939, at forty-six.
His historian is careful about the claim usually made for him, and we should be too: while he may not have invented the cryptic clue, he was the first compiler to use cryptic clues entirely. That's a real distinction and a smaller one than "inventor".
The rule of fair play
What Torquemada didn't supply was a code of conduct. His puzzles were brilliant and, by later standards, occasionally lawless.
The correction came from two men. Ximenes - Derrick Macnutt, his successor at the Observer - codified what a clue owed its solver. And Afrit, A. F. Ritchie, Prebendary of Wells, gave the tradition its single most quoted sentence, in Armchair Crosswords of 1949:
"I need not mean what I say, but I must say what I mean."
That's the entire ethical basis of the cryptic in eleven words. The surface reading may deceive you completely; the actual instruction must be literally, defensibly present. A clue may mislead. It may not lie.
(The line circulates far more often in the second person - "You need not mean what you say…" - but the first-person form is what Ximenes quotes, twice, and it's the earlier.)
Two grids, two philosophies, and a disagreement about what a setter owes a solver that has never been resolved because it was never really a disagreement - just two reasonable answers to the same question, taken to their conclusions on opposite sides of an ocean.