DOWNLOADABLE PUZZLE: Unthemely #95 (PUZ) (PDF)
* * *
While working on this Unthemely puzzle, I started thinking about traditional crossword specifications and the rationale behind them. Some specs serve a practical purpose. Checking all letters in a grid, for example, reduces the chances of a solver getting stuck on a single unknown answer. Some specs reflect the standards used in adjacent journalistic and entertainment practices. The rules about obscure and indelicate vocabulary fall in this category. Some specs are simply aesthetic choices invented for the medium. The rotational symmetry of black squares is such a choice.
Rather than putting crosswords rules in specific categories, suppose we rank them on a “rationale” scale of 0 to 10, where 0 represents practical/derivative and 10 represents arbitrary. The “No two-letter words” rule might rank a 3 on this scale, while “No more than 72 entries in a themeless grid” might be an 8 or 9. I have a question based on this hypothetical scale: where would you rank the rule that discourages or prohibits the duplication of words or word forms (e.g. EATS and ATECROW) in the same grid? Remember that a ranking reflects how much the rule serves a practical purpose or represents an artistic standard that extends outside the world of crossword puzzles.
life as a puzzle 
Not sure of the number, but that definitely seems like an example of elegance within the artistic medium — similar to, say, “don’t repeat the same rhyming word in two lines of a limerick”. The justification of “we don’t allow it because solvers would get confused” is basically only true because solvers have been trained to expect it. (All that said, of course, I completely support the rule, as I suspect you do.)
True. I support many crossword rules that I would rank on opposite ends of the rationale scale. The comparison with limerick rhyming words is interesting but I doubt poetic tradition was a specific inspiration for the crossword rule. I suspect it’s closer to the way that crossword fans want to celebrate the variety of the language by valuing entries with low-frequency letters. It’s a standard of elegance that was arbitrarily established somewhere along the way.
I don’t know how I feel about it in crosswords specifically, but when I’m gridding a variety puzzle and I seem to have a lot of the same prefixes or suffixes with similar overlaps between them, it just feels less interesting to me. Maybe solvers don’t mind, but I don’t want to use the same wordplay multiple times in one puzzle, unless that’s the point of the puzzle. Now that I try to place it on the practical/arbitrary scale, I’m trying to decide what “practical” means. Checked letters make a puzzle more tractable, but really, given the point of most puzzles, being tractable is just one way to keep puzzles from being frustrating, that is, not entertaining. Some duplications are annoying, and others are entertaining. Uh…what was the question again?
Maybe, instead of practical/arbitrary, the scale should be empirical/instinctive. Some rules of puzzlecraft are established for reasons that are easy to explain or reference. Others are not easy to explain but just “feel” right, elegant, more interesting, etc.
Okay, I’ll add another one — partials. I think wanting to avoid partials is pure crossword-person snobbery. As a solver I never minded them; in fact, I liked them because they were easy for me. I wonder where others would put them on your 1-10 scale.
I don’t know of a mainstream outlet that prohibits all partials. I guess a puzzlemaker who adopts a “no partials” rule could justify a low rating based on the standard syntax in other forms of communication. The typical “no partials more than five letters” rule rates close to 10.
I agree that “no partials more than five letters” is arbitrary, but I suspect many other puzzlemakers would disagree.