I am not quite sure when I first had the idea of a crossword with double clues leading to two possible gridfills, one of them to be submitted and the other discarded. The solver obviously had to be able to determine which was which and I went for two crucial messages (hence the title Double Cross), one saying THESE ANSWERS SHOULD BE USED and the other PLEASE IGNORE THESE ANSWERS (I liked the way THESE ANSWERS occupied one diagonal in one case, the other diagonal in the other). But how to deal with the solver who found the ‘wrong’ gridfill and decided that all he had to do was submit a blank grid (PLEASE IGNORE THESE ANSWERS!)? The puzzle needed an additional ‘cautionary message’ (an excellent phrase suggested by the Listener editors) and that surely had to come from the clues, and I went for ‘rogue’ letters. In order to continue the ‘double’ theme, thereby reinforcing the title, I decided that in each ‘clue’ (i) it might be the right clue before the wrong one or vice versa, (ii) the rogue letter might appear in either, and (iii) the rogue letter might need to be either removed or restored: so there were to be 2 × 2 × 2 = 8 different types (to allow both myself and the editors to keep track of things, I had to devise a notation: 1/2 meant the right/wrong clue, and the eight types were 1+2, 1–2, 12+, 12–, 21+, 21–, 2+1, 2–1), and I further decided that there should be the same number of each type, a fairly typical Radix decision, as I hope some of my few fans will admit. So the number of clues had to be a multiple of 8, which in effect meant 40 (32 is hopeless in a 12 × 12 grid, and 48 is much too great). So the cautionary message had to be 40 letters long, and after a lot of this and that I came up with YOU MAY NEED TO CONSIDER THE OTHER CLUES INSTEAD (again, I liked the way YOU MAY NEED TO CONSIDER came from the across clues, the rest of the message from the downs).

Now I could write my preamble (Radix and the consortia to which I belong decide on the preamble, more or less fully, before turning to the grid and fill, let alone the clues!), which read:

Each ‘clue’ is actually two clues side by side, with a rogue letter either inserted or removed somewhere along its length, twenty of each; after a few obvious case adjustments, repairs always leave proper words behind (one of them not in Chambers), but the surface sense does usually suffer. Only one of the two clues need be solved initially, but the rogue letters spell out advice that may prove helpful. Solvers must finally highlight a thematic instruction and write the alternative beneath the grid. The Chambers Dictionary (2008) is the primary reference.

The editors altered this a bit, some of which (‘cautionary message’, for example) I liked.

So now to the grid. There had to be 20 across slots, 20 downs. And the unching had to be Radixean — of course! Radix follows Ximenean principles (4/5 letters = 1 unch, 6/7 letters = 1/2 unches, 8/9 letters = 2/3 unches, 10/11/12 letters = 3/4 unches), but insists further that a shorter slot should never have more unches than a longer slot.

So now to the two fills. No great problem there, although it was not easy to find two fills that I was more or less satisfied with (not too many ‘bad’ words — plurals and the like).

So now to the clueing, which I found incredibly difficult, subject as it was to my 8 × 5 requirement, especially since I further required that the 8 types should be distributed as evenly as possible among the forty clues. The difficulty, of course, is how to marry two clues so as to produce one sensible ‘clue’. For instance, how to clue AHAB and OLLA: my solution, after a lot of head-scratching, was: “Hello Sailor!” — Captain | [u]RN in variety entertainment [AH + AB, Two meanings]. But what about KILERG and ENJOIN, what about GRISEOUS and BLESSING? After writing about ten ‘clues’, I gave up in despair, and only returned to the task, about two years later, having been encouraged by the brilliance of a fellow setter in this area. I worked and worked and worked on every single clue, one after another, and finished with forty that I felt were 100% sound, and so was somewhat disappointed that the editors thought it necessary to fiddle with so many.

I was reminded of Don Manley’s “Ten tips for editors” in his Crossword Manual, number 8 of which reads as follows:

Faced with a clue, you have three options: (i) leave it, (ii) amend it slightly, or (iii) rewrite it. If a clue is sound and sensible (and hasn’t appeared recently), you are best to leave it (option (i)) unless you can think of a small amendment which will add a touch of gloss (option (ii)); …. Only make the change if (a) it adds finesse or (b) it renders an unsound clue sound. Go for option (iii) if the clue is grossly unsound or doesn’t make sense. …

In my case, the editors never went for option (iii), which would have been unwarranted, but in my opinion went for option (ii) a bit too often.

P.S. In the event, nobody submitted a blank grid, but a dozen or so submitted the ‘wrong’ grid. So the two instructions and the cautionary message proved inadequate, for which I apologize.

Radix, 01/05/2010