Although this book is famous for being among Sellars' most difficult, I actually found it (at times) far easier to get into than his more-beloved Empiricism and the Philosophy of Mind, although this is perhaps due to the fact that I'm more familiar with the philosophy of science than of perception. There is certainly a lot here that left me rather mystified (lots of formalisms the utility of which only rarely becomes obvious), but when the book shines, it shines. It in many ways resembles (preempts) Michael Friedman's Dynamics of Reason, another Kant-inflected book about continuity / convergent realism in science. I'm not really sure how to summarize a work like this, but the passages about the relationship between successive theories (in Chapter 5) were quite illuminating, as was Sellars' insistence (in Chapter 2) that "[the fact that] things-in-themselves have categorical form as represented... [in no way implies that] they cannot exist in themselves... it can scarcely be overemphasized that the difficulty Kant finds with things-in-themselves is that... our conception of them is empty—not that it is incoherent." (On the latter point, I remember being quite puzzled in an undergraduate seminar by the professor's seeming implication to the contrary!)
In a way, it reminds me of that Bolaño quote, about the pharmaceutical clerk who prefers the "perfect sparring matches" of great writers more than their messy attempts to confront what really matters (or something to that effect). For make no mistake: this book is all over the place. Chapter 6, for example, begins with a series of summary comments regarding material which, to my memory, was not covered in the preceding chapters. But in the end there is something quite admirable about it.
I heard something about Sellars being good once. What's the point of the book, besides that maybe the thing in itself isn't incoherent? (I thought the difficulty wasn't the emptiness of the thing in itself, but the fact that it has to both cause phenomena and exist separately from the causal/phenomenal realm.)
Probably the safest general characterization of Sellars' project here is an attempt to give a (Kant-inflected) defense of scientific realism—Brandom memorably characterizes it as something like "the noumenon is replaced with the content of a mature scientific theory". This involves a defense of the Kantian "manifold of sense" as against the empiricist / naive realist account, wherein we encounter objects in the world. Drawing on Peirce, Sellars situates the adequacy of our given conceptual scheme (CSO, for "conceptual scheme ours") in terms of its relationship to an ideal / "final" conceptual scheme (CSP, for "conceptual scheme Peirce"). There's a lost of philosophy of language stuff about the relationship between propositions of the form "X stands for Y", "X represents Y", etc., but honestly I read that stuff and it kind of goes over my head, at least on my first pass. Also some stuff about how our "theory" of mental states is formally analogous to the way that scientific theories are formulated (as an expression of some invisible force / cause underwriting the observable phenomena of speech acts), but that's really mostly drawing on EPM. Honestly it's funny reading this book, because it almost feels as if EPM were subsumed directly into it. As a result, "Philosophy and the Scientific Image of Man" and "Empiricism and the Philosophy of Mind" are probably better places to start with him, since this book kind of incorporates a *lot* of his material into a single (gloriously messy) package. Sellars is generally acknowledged to be very, very good, and in my experience he's definitely worth reading. He's (nearly) unique among the 20th century analytics in that he has real systematic ambitions, and is quite engaged in the history of philosophy. This is my first read of the text, and honestly I'm using it mostly to gather ideas for my own work, rather than in an attempt to develop a super clear exegetical interpretation of the text. Willem DeVries' "Wilfrid Sellars" is supposedly a good exegetical overview of his work, and John McDowell is supposed to have done good work on him too—as, of course, many others have. The only scholar I've heard—well, not negative, but, we'll say "conditional" things about is Brandom; from what I gather, his book on Sellars is almost a bit Deleuzian, in that he's thinking alongside and against Sellars in order to better articulate his own project. (Incidentally, I think there's some kind of sympathy between Sellars and Deleuze—there's even a small literature to that effect—but that's neither here nor there...) Hopefully that's helpful, happy to ramble further!
Interesting. Happy to hear he has systemic ambitions, since no small number of analytic philosophers remain silent on that matter, I think because they find it naive. Engagement in the history of philosophy is good also since the analytics are generally very insular. I haven't read Deleuze. If you have the time: If "the noumenon" is replaced with the "content of a mature scientific theory"--how is this theory selected, can it change over time, and what surety do we have for it against skepticism? What more precisely is a conceptual scheme? (is it related to Kant's schemata?) How does Sellars fit with your own philosophy?
The key thing for Sellars is that scientific theories converge over time towards a more adequate representation of the world, so change is inherent to his model. My instinct is to say that skepticism isn't really a live problem for this sort of avowedly fallibilist picture; regardless of whether or not this is "the right theory", it *works*, it descends from theories which worked (albeit to a lesser extent), and we can continue to refine it on the basis of new information. That's more coming from Peirce (esp, iirc, "The Fixation of Belief") than from Sellars, but Perice has a pivotal role to play in Sellars' text, so it seems appropriate. On the relationship to Kant's schemata, I'd need to go back and review Kant a bit. But a conceptual scheme is basically just the set of beliefs / theories that we have, or the propositions which we can assent to, or something like that. It's the framework, or paradigm, or web of belief, out of which we're operating. As far as my own philosophy is concerned I've been writing a paper on how the Kuhnian perspective on theory change accidentally enables pseudoscience (since, if science is revolutionary in the way that Kuhn presents it, we can't rule much out from the get go), and how a convergent realist picture is better able to explain the unjustifiedness of certain theories by showing how those theories contravene the general convergent trend of scientific inquiry, drawing on some of the technical materials in Sellars, Friedman, and Peirce. Or, to put it another way: convergent realism gives us the resources to articulate rational norms which persist *across* paradigms, rather than being limited to the paradigms themselves. I also compare some recent computational models of Kuhnian theory change with my own model, using genetic programming, and show how the dynamics of my model are structurally analogous to the dynamics which Sellars presents in Science and Metaphysics, with the idea being that, if I've presented a good computational model of how learning works, and that model is closely related to Sellars' model, then the strength of the computational model affords at least *some* credibility to the Sellarsian picture. Which is maybe a sort of "experimental" or "fallibilist" way of doing metaphysics.
I can't say whether it works or not without looking into it further, but what I can tentatively say is that the scientific/approximatory approach to metaphysics is something I was completely convinced by around 2-3 years ago, and which I now don't pay much attention to at all. There is a little bit of vagueness to me in this notion of a scientific theory "working" and "representing" the world. If the "working" of a scientific theory means only that it can predict successions of phenomena it doesn't go beyond Hume. If someone wants to go beyond Hume and grasp at metaphysics, like Kant did, he should have a very novel and ingenious solution for the problems that Kant's system faced, and I don't think the switch from paradigms to "rational norms which persist across paradigms" is that. But again I am unfamiliar with Sellars, and with Peirce beyond his mathematical works.