I should write at the outset that I don't pretend to have a handy guide to how to read Adam Roberts. This is mostly a response to Paul Kincaid's blog post here.
Paul Kincaid describes an interpretive tool that may work for his own reading: Roberts writes Menippean satire. There's something to be said for that. Certain elements of the satire are there: fragmented narrative, the rapid movement between styles, the use of the picaresque, satirical commentary. Kincaid seems to me to be certainly right that many of Roberts' books work as satires of SF. (See, e.g., my piece here on Splinter.) But. The concept of the Menippean satire comes with a lot of baggage that doesn't seem to apply, and doesn't include a good deal of what I see as being there. The concept may work perfectly well for Paul Kincaid, who states upfront that he has trouble reading Roberts, but as a fan writer who's written a good deal about Roberts' work, it's not how I think he is best read.
Roberts' works seem to me to be best described as experimental novels. They would be avant-garde if there was now any literary garde to be in the avant of. Kincaid writes at the start of his post that "I don’t believe the set up, I don’t believe the characters, the plots seem an exercise in artificiality." Well, yes, sometimes they do. Adam Roberts, if I can resort to biographical criticism, is a professor and historian of SF who has read a tremendous number of SF works. The Ideal Author of his books, to revert to how they appear to me to be written, is not someone who wants to write a seamless set-up, believable characters, and realistic plots. Those elements of writerly technique have long since been turned into routine. Although I admire the craftsperson who can successfully carry out this routine, I'm really more interested in attempts to do something else.
What is the "something else" that Roberts attempts? His novels seem to me to be characterized by a high degree of formal structure coupled with underdetermination of what that structure means.
In Swiftly, in another passage that Kincaid had trouble with, we see Eleanor before and after a big gap in her described life, and she appears to be almost two different people in the two sections of the book. One could write a middle section of the book that explains the transition from one Eleanor to the other. But that would remove what Roberts is calling attention to: the gap itself, which the reader can and must fill in with their own narrative. "Blood in the gutter" is a comic-book critical phrase referring to the space between one comic book panel and the next, which the reader must fill in out of their imagination. By making the gaps in action small enough so that readers ideally never notice them, authors create "realistic characters". But what they are doing is pretending that they are doing the work that they are actually trying to get the reader to do. Swiftly is a formal balance with a beginning and an afterwards around a gap, and wouldn't work without it.
Does this kind of thing always work? No. Sometimes Roberts may attempt a bit too much at once. Yellow Blue Tibia might have had a bit less of the (yes, distracting) pseudo-Aspergers sidekick character if it hadn't been "how about a metafiction about Soviet SF coming true" mixed with "what would an action movie hero look like if he was old and plunked into a pseudorealistic setting". But experiments are experiments: some work and some don't. Literary experiments are better than scientific ones because literary experiments work for some people and don't work for others. Jonathan McCalmont wrote some excellent criticism of New Model Army that takes the same changes in viewpoint that Paul Kincaid notes and turns them into a sort of satire of SF yet again, "a cultural blueprint for the entire genre."
Do I think that Roberts intended to write the book that McCalmont read? No, not exactly. They are, to repeat myself, books with strong formal elements coupled with underdetermination of what those elements mean. In other words, they're wonderful toys, full of moving parts that don't do obvious things, made to order for a critic to play with and come up with a reading. I can understand why a bit of puzzlement is common when reading Roberts. If New Model Army is supposed to sit on the bookshelf shelves next to, oh, Lois Bujold, or Vernor Vinge, then why doesn't it keep the same viewpoint throughout? Why the scene "which feels like an intrusion from a different work altogether," to quote Kincaid? Write the thing in an accepted military-SF style throughout and then it'll be normal. But it's really supposed to sit on the shelves that hold, say, Aldiss' Report on Probability A.
Kincaid comes close to picking up on what I find interesting about Roberts' work. But a satire of the kind that Kincaid envisions is organized around a purpose. That purpose doesn't seem to me to be at the heart of these works. These are writerly artifacts, and while some of them work as satires -- Roberts writes what he knows, and what he knows is SF -- they do not seem to me to pick out a direction and then set out to guide the reader along it.