It is easy to see why this is considered an important book. Dennett synthesizes and moves fluidly through a vast scientific literature to come up with a broad theory of consciousness. Though he is overreliant on neologisms and metaphors, the contrasts that he draws between the Cartesian Theater and his Multiple Drafts model provide a clear structure to the book. The preamble about heterophenomenology being about interpretation of a text concerning a fictional world---and the whole “narrative” metaphor in general, with the Multiple Drafts, Joycean machine, etc.---is a bit overblown and probably could’ve been excised.
Dennett is irritatingly cagey at times, promising to deliver on delayed explanations and only half-fulfilling them. He is a gifted writer who rarely veers into inscrutability, but his aversion to arguments, as opposed to his preferred style of folksy discussion and intuition pumps, can be a detriment, especially with the more involved passages on experiments like the color phi phenomenon, Libet’s timing experiments, etc. His penchant for metaphor can be irritatingly self-referential, which makes some passages opaque (consciousness is a “von Neumann-esque virtual machine implemented by a pandemonium of demons”) but he does take time to explain what he means.
The usual criticism of this book---that it really isn’t about consciousness — is something that Dennett acknowledges in the book itself. I have mixed feelings about this. On one hand, I commend him for actually trying to come up with a detailed theory instead of just wringing his hand and saying that it is impossible or outside our ken, like most dualists do. On the other hand, I do wonder if he is missing something quite important. It is clear that Dennett is a naturalist through and through, but his commitment to a certain methodology — i.e., that of science---is coloring his metaphysical commitments. Is he putting the epistemic cart before the metaphysical horse? Maybe, but it’s a dazzling book anyway.