And it is. It's basically an explanation of the grammar of comics: how the frames are used; how time is perceived; how motion is conveyed; how the quality and realism of the art affects reader identification; and so on. And it's done in the manner of a comic, with some nicely presented conceits as a result. McCloud's avatar is used to good effect to convey his approach. Not without irony, though, since this is a book that says, "this is why this device is having this effect," not, "if you see this on the page, read it this way." Because, obviously, if you needed the latter, you wouldn't be able to read this book.
He goes into much greater detail than I'd care for, in some regards. For example, much is made of the inability to see anything other than a face, in the traditional circle-line-two-dots arrangement of a simple cartoon. For another, he develops theories of how pictures (received information) and words (perceived information) interact, and how both should be simplified towards a common level (away from their own rarified ideals) if they're to work in conjunction. And then he goes into less detail than I'd like on other areas. Some good stuff on the different panel-to-panel transistion types, and a whole chapter on each of time and frame usage, but there were comments on the use or otherwise of frames, in one paper on Gaiman's The Tempest that I thought came from McCloud, but which aren't there - the paper author has extrapolated futher.
It's an impressive analysis, though, taking in a lot of history (Mayan and Egyptian, for example), plus US, European and Japanese comics for different evolutionary approaches.
But it doesn't take it far enough. I'm going to have to read Eisner's book - well, books - too, aren't I?