In this post over at Literal-Minded (another linguistics blog—yay!), Neal Whitman talks about the oddness of the answer-questions and answers-in-the-form-of-a-question on Jeopardy. He's on to something, but I think there's additional weird stuff going on in Jeopardy clues.
Neal is right to point out that they're sometimes pretty lax about whether the contestant's response is really a question that would elicit the clue up on the board as an answer. If the clue is:
(1) This man was the first President of the United States.
...they'll accept a response like "Who was George Washington?" even though (1) isn't a very felicitous answer to that question (in fact, I don't think it's a felicitous answer to any question, unless the answerer is also pointing at something). Neal also points out that they're really looking for one particular response, not just any question to which the clue is a correct answer—otherwise, contestants could always respond "What is the current clue in this Jeopardy game?". A commenter over at Literal-Minded points out correctly that the category of the clue constrains the possible answer-space, but I think the clue-writers at Jeopardy are also doing something syntactically surprising to hint at the right answer.
Consider this clue, slightly different from (1):
(2) He was the first President of the United States.
As with (1), "Who was George Washington?" would be a correct response to this clue, and they make up a more felicitous question-answer pair when reversed, too. However, there are also a large (infinite?) number of other correct responses like "Why is George Washington famous?" or "Why is Washington's face on the one dollar bill?" This clearly won't do.
To reduce the number of possible responses, the clue-writers often give extra information in the clue, and mark it with a demonstrative. Check out this page, which has the clues from a Jeopardy show from 2002. Note the italicized parts of the following examples:
(3) Automaton is sometimes just a fancy name for one of these mechanical beings.
(4) This species of bear is the best swimmer
(5) There's no need to be ashamed of this common mental illness, even if it's alternating with mania
In all of these cases, as with (1), the correct responses ("What are robots?", "What is the polar bear?", and "What is depression?") make pretty unnatural-sounding question-and-answer pairs when you put them before the corresponding clue (try it). I think the purpose of the demonstratives is to focus the attention of the contestants on the constituent in the clue that their response needs to supply. In the game I linked to, the majority of the clues include a demonstrative phrase used this way, with the exceptions of the category "A Beautiful 'Ind'" (where the clue is a definition of the correct response), the category "Classic Drama Series Episodes", some of the daily doubles, and few cases where a pronoun without an antecedent is used instead of a demonstrative.
What the Jeopardy writers have done is invented a way to do in situ Wh-questions in English using demonstratives to mark the Wh-element, instead of the usual pattern of moving the Wh-element to the front of a constituent question. Part of what makes Jeopardy fun, I think, is this extra bit of grammar that contestants and viewers have to acquire in order to play—the way the writers phrase clues in order to elicit particular responses is a narrow, Jeopardy-specific dialect of English, and learning it gives us access to the game.
ObLingSF: Here's a particularly unnatural-sounding response and clue from the April 30, 2002 game, in the category "Classic Drama Series Episodes":
Q: What is "The Twilight Zone"? A: "To Serve Man"
One of these days, I'll get around to writing about that (linguistics-related) episode and the Damon Knight story it's based on.
[Now playing: "Yin and Yang (The Flowerpot Man)" by Love & Rockets]