I put the last letters into yesterday’s Listener puzzle around 12:30 am this morning and then fell asleep. It’s called “Printer’s Devilry” and it actually felt to me like it was too complicated, that there was too much going on — and I’m not one to say that lightly when it comes to a puzzle. There are three kinds of clues: some contain one misprinted letter (you have to find out which one is misprinted and what it should be); in other clues, a letter has to be dropped from the answer wherever it appears before you put it in the grid (in other words, REFERENCE might be entered as REERENCE or EFEENCE or RFRNC and so on, depending on which letter you had to drop); in still others, the answer had to be encoded before being entered. Individually these are all interesting twists to put into a puzzle, but the combination was less fun to solve than I thought it was going to be. And in spite of several very enjoyable discoveries along the way: Finally figuring out the two-word phrase that the puzzle is organized around was a pleasant surprise, and the final grid turns out to be more orderly that it first appears — really a remarkable piece of construction, in fact.

I was mostly frustrated by the misprinted clues. I’ve most often seen this gimmick used in cryptic crosswords where the instructions said that the misprint was always in the definition part of the clue, and it has seemed to me that this is the most satisfying way to use it. (Thus the definition part of the clue might be, oh, “word from a quilter”, and the answer might turn out to be UNCLE and you’d have to figure out from the rest of the clue that “quilter” should be “quitter”.) In this puzzle, though, the misprinted word could be anywhere in the clue. In one case it was the anagram indicator that was misprinted, and I could think of at least three possible words that the “correct” (that is, before the misprint) word could have been. In another case it was in a two-letter word and there are two clearly valid possibilities for the correct word. Since you’re supposed to be keeping track of the changed letters and using them to spell out a message, it turns out that the only way you can actually figure out what the correct words are supposed to be is to guess the message without the help of those clues and then back-solve from the answer. This seems inelegant to me.

There were also cases where the “correct” wording of the clue seemed awfully tenuous and contrived to me. Probably because of the restrictions imposed by that very message spelled out by the changed letters. Both the original and the changed letters were involved, which may have been too much of a constraint for the constructor. It’s one thing to come up with an interesting clue for UNCLE using the misprint gimmick; quite another when because of all the other constraints you’ve worked into the puzzle, the clue must involve specifically, say, the letter N being misprinted as an I. Going to be tricky enough to find any valid clue at all under those circumstances, let alone a really satisfying one.

The rest of the puzzle, though, was mostly sharp and fun to solve. Learned a couple of bizarre new words, too.