Go forth and find readings?
Undine's post reminded me of this article in the NYT about an experiment in secondary school teaching: allowing students to pick some/all of their readings for the term. (A poster at Undine's noted that this was not, in fact, a new idea.) Could you implement a similar program at the undergraduate level?
I don't think this would work in any introductory-level survey course. Part of this has to do, no doubt, with my pedagogical bias: I firmly believe that students need to know "the canon" (insert necessary qualifiers here about the shifting makeup of "the canon" over time) before they engage in any canon-busting behavior. A survey is a kind of unfinished map. It can't include everything (or even most of something), but it can at least give students some basic ideas about who was who and what was what. Ergo, the students need a pre-set syllabus...which is not to say that you can't discuss how both your syllabus and your chosen anthology came to be.
I do think this would work nicely in any course where content takes second place to skills. Our Introduction to Literary Analysis course, for example, could probably accomodate a free-for-all with some restrictions (e.g., letting students choose any works they wanted, but requiring a specific range of genres). If you specifically wanted students to learn how to think and write about genre conventions, for example, then popular fiction and/or film would actually work rather well. (I'm doing the nonfiction version of this approach in freshman comp: we have textbooks for academic research and style, but the students are spending their semester on their own research projects. Yes, there are exercises designed to make sure that they're doing some reading...)
A modified version of this approach might make for an interesting upper-division seminar: given a particular topic (e.g., "Historical Fiction in the Victorian Period"), the students and instructor would develop the syllabus together. There have already been some interesting experiments with such "collaborative" course construction. For example, you'd spend a couple of weeks showing them how to work through the necessary literary histories and online databases, followed by a discussion of goals and priorities (what do we want to know by the end of the semester? do we want to focus on a specific geographical region?), not to mention syllabus logistics (er, wait, is that even in print?). And, ultimately, a syllabus. This would also work well as a graduate course, especially in a more advanced seminar.
On a related note, I've always thought it would be interesting to teach a graduate course in which students annotated and introduced either a completely non-canonical work or a "minor" work by a major author.