Okay, so here I am.
First the quote:
I was a bit surprised at b., and sure enough you made a quotation mistake (as to amazon's sample text). b. should read:
He looked over at his wife. She looked so unhappy, he thought, almost sick. He wondered what to say.If we're nitpicking terminology, that's important, as otherwise Wood would have used the term "reported speech" differently than anyone in linguistics, which would be odd.
b. as you quoted it, is a punctuation variant of a. It also exists, quite frequently, in mainstream fiction, while genre has taken to using italics (even with attribution, which is a recent development). Direct thoughts with quotationmarks were common once, but today they're almost extinct, coming up mostly in fairy tales and other books for small children.
As far as I can tell from the amazon excerpt I read, Wood does not use the term "3rd person omniscient" in the way that we're used to on message boards. He approaches PoV via reliability:
Wood then goes on to deconstruct that difference, saying that first person narrators aren't necessarily (or even primarily) unreliable, and third person narrators aren't always reliable. So forget 3rd person omniscient vs. 3rd person limited. For Wood, there's only 3rd person and 1st person; and 3rd person is associated to omniscience. The preview ends to early for me to tell, but it looks like he's analysing 3rd person according to how the narrator subverts his omniscience.
Originally Posted by Wood
This is a very different approach from what we're used to on writing boards. What's important in the context of the quote is: "omniscience" is an aspect of "3rd person narration". What he's basically talking about - in our more familiar terms - is the creation of third limited; but his take is different: rather than the "author" "hiding" behind the view-point character, the "author" is "merging" with the view-point character.
Note that Wood is - at least in the beginning - conflating the omniscient narrator with the author:
Humbert Humbert is the first person narrator in Lolita.
Originally Posted by Wood
He probably does so for convenience' sake. (At least I hope so, as this sort of conflation is pretty much out of fashion for ages, and it wouldn't work at all for the omniscient narrator in Jonothan Strange, to give just one example.)
What we're lumping together under point-of-view are actually two distinct things:
1. Focus: Through whose eyes do we see? Wose evaluation do we get?
2. Presentation: Whose language do we get? Who do we attribute the words to?
You'll notice that focus stays the same from 1.a to 1.c, but that the presentation keeps approaching that of the character. "he thought", "he wondered" are words we attribute to the narrator. 1.c has none of those left. The only hints at the narrator in presentation are: pronouns ("he" instead of "I") and tense (past instead of present).
The sentence "He looked over at his wife," though, is neutral, PoV-wise. It's a representation of his actions. You don't think "I do this/I do that", you're just doing it. Generally, these words should be attributed to the narrator (both focus and representation-wise), but they don't break PoV; i.e. we don't refocus onto the narrators PoV, instead we focus on the happenings.
This is why I don't think the familiar three point-of-view model is very useful. It makes sense in theory, but in practice texts are more complex and can fall anywhere between the categories.
Basically, Wood's free indirect speech ("free indirect discourse" in narratology) is the key technique in what's usually called "3rd limited" in creative writing seminars and on webs; third limited is characterised by this method - other than PoV-neutral text, nothing can occur in limited than this. (This doesn't mean squat for writers - it just means that what they've written would be analysed under a different name.) Omniscient (in seminar-speak), however, also has access to that method.
So now we come to the difference between 1st person and 3rd limited: the first person narrator knows he's telling the story; the viewpoint character in 3rd limited doesn't. What this means is that the view-point character close presentation and focus in 3rd limited stays very close to the experiental level, reducing narration to a minimum. Thoughts are represented as they occur, and actions are described. Many techniques that both 3rd omni and 1st person have (such as flashbacks and comments) don't exist in third-limited.
In third person such extra-experiental would automatically be attributed to the narrator in focus. For example:
First person: I spent all my allowance on booze. I was so young and foolish.
Third person: He spent all his allowance on booze. He was so young and foolish.
See? In first person, the narrator (old) comments on himself (young). In third person, the narrator ("author surrogate") comments on the character. See? There is no way that third limited can comment from a later PoV. All those sentences are different from either plain description (of action or perception) or free indirect speech (character thought in close presentation).
This is why so many people argue that 3rd limited is the "most intimate" of PoVs. But these people also usually are too strict with "no PoV-changes"; 3rd limited is still a subtype of 3rd omni, and PoV modulation is far easier to do well than they think. The problem, I suspect, is that often the terms take on a life of their own - as if "3rd person omni" and "3rd person limited" were actual points of view rather than crude terms used to think about very complex (= real) PoV constellations in actual texts.
So, to sum up the main point, yes, I think "free indirect speech" (I prefer "discourse" over "speech", but let's not be picky) is a useful concept. But please bear in mind that (a) Wood does not speak about what you would probably call 3rd omniscient, and (b) he orders the field quite differently from what you're probably used to, focussing on reliability as the key term.
As always, this is fiendishly complicated, so I hope it makes sense.