There’s a difference between the allegorical or symbolic literalization within the work of, for instance, Kafka and Borges, and the absolute, take-it-for-what-it-is literalization of most science fiction, which is one reason people who can read Kafka with no problem get stuck when trying to read genre SF. (The danger for people who read nothing but SF is a tendency to become dulled to metaphor as metaphor, because SF readers become so adept at suspending the faith in metaphor that Farah mentions. But that’s another subject entirely…)
My (superficial) response to Matt is below: Well put. Genres do overlap a lot; still I don’t think Kafka would seriously consider the idea that the castle could literally happen the way he suggested it. Actually, a better example of “Kafka sci- fi” is Amerika, a truly bizarre take on America. (which has next-to-nothing to do with the country where we live). He didn’t even try to make it plausible. Contrast that with speculative lit which frequently goes to great effort to produce a plausible effect.
Kelley’s characters are probably the first in scifi I’ve encountered that really linger in my memory (which may be saying more about my ignorance of the genre than the genre itself).
The literary v. scifi dichotomy probably is more useful for talking about behind-the-scenes (how the inspiration came about) than the story’s final effect. In plot-oriented or even a story of ideas, characterization arises from plot, philosophies, allegory, etc. In typical “mainstream novels,” you start with the voice and then bestow a situation. From the reader’s point of view this might not really matter, just as we don’t really care if the Pixar characters were created by a novelist or a graphics designer playing around with his technology.
That said, the “time capsule” aspect to novels or films does have value; we read Updike’s Rabbit is Rich, just as much for literary pleasure as the chance to reexperience the 1970’s. Or to imagine what life in the 1970’s used to be like.