I was watching an old episode of Seinfeld the other night, and a line of dialogue caught my attention: "That's the guy I told where the elevator was." It's not ungrammatical for me, but it's marked in some way that made my analytical skills perk up. According to current syntactic theory, this type of sentence is created by movement from one original, underlying position to a different surface position. You don't have to take this literally as movement; many syntacticians take "movement" as a relation rather than an actual move from one position to another. But the key is that a clause like "where the elevator was" is assumed to be in some way structurally the same as "the elevator was there". In some analyses of relative clauses, the same thing is true for nouns described by relative clauses, so that the DP "the guy" comes out of the CP "I told <the guy> where the elevator was". So the so-called D-structure (don't read any theoretical assumptions into that, I'm just using it as a convenient label) of "That's the guy I told where the elevator was" would be something like "That's I told the guy the elevator was where", changed by movement into "That's the guy I told <t> where the elevator was <t>".
But why did I find this remarkable? I'm fairly certain I use such structures. Before I analyzed the sentence, I thought it might involve some sort of movement where one constituent crosses over the trace of another (which, at least in Relativized Minimality, wouldn't necessarily involve a violation, but might be more marked in some way), but this isn't the case. The "the guy" movement occurs entirely in the matrix clause, and the "where" movement is entirely limited to the subordinate clause. So there's no crossing involved, just regular movement in two separate clauses. I tested out some similar, simpler examples, and found out that I find "I told him where the elevator was" completely unremarkable, but "That's the guy I told to clean up the mess" (very) slightly remarkable. (Side note: can we really have judgments about such minute differences? Maybe not everyone, but I've been dubbed "most sensitive speaker of English" at Rutgers, so I claim the right to make such judgments. I'll happily give different numerical scores to subtly different subjacency violations.) So what, I don't like relative clauses? It's not that, because "That's the guy I told" is also completely unremarkable. I think it's some very slight underlying preference for traces to be at the end of a clause. This would make questions unremarkable ("Who did you talk to <t>?"), as well as simple relative clauses ("That's the man I talked to <t>"). But more complex relative clauses with material after the trace make some small part of my language faculty slightly uncomfortable.