I have some issues with the existing theories:
- problems with Scott McCloud's panel transitions - I love Understanding Comics. It's a great primer, and makes a great case for taking comics seriously as a medium. The theory of panel transitions Mr. McCloud gives in the chapter on Closure is descriptive enough to be useful. My only problem with it is that it's not flexible enough to be universally applicable. Just off of the top of my head, I can think of a few cases where it either fails to account for something or is incomplete: the parallel sequences of newsstand and pirate comic in Watchmen, the Kindly Ones switching eerily switching positions in an early issue of Sandman, and the parallel pages-within-pages in the "Two Lives" issue of The Spirit (which Eisner talks about in Comics & Sequential Art). Also, the Time chapter of UC gives an example of how a single panel that seems to portray several points in time at once would qualify as "comics" if some gutters were added—I think that kludge could be avoided if primacy was given to connections between images rather than specifically panels. Occam's razor!
- problems with Neil Cohn's deep structure theory - Basically that it's only really interesting to people who study cognitive science. It may be possible to use it to analyse a comic, but I'm not sure what that would tell you about the comic, and I don't think it can really help in creation. Also, it tries to be really strict, but it leaves a big loophole with "transformations". After he spoke at Comic-Con, in a presentation where he asserted among other things that visual language does not allow a sequence to "split", I brought up an old Silly Cone V by scrubbo in which the comic splits into two different columns, each representing a timeline split by a choice (probably my favorite SCV, BTW). Neil's response was "oh, a transformation allows that", which (a) explains absolutely nothing, and (b) means that all of that stuff about inability to split in "deep structure" irrelevant if it can just be "transformed" to fit.
- problems with Neil Cohn's web diagrams theory - None, actually, except that Cohn seems to have abandoned it prematurely in my view. This is what I want to expand on and unify with Kress & van Leeuwen's "grammar of visual design".