I want to write some in depth posts about the material in chapter 4, but I have precious little background in the philosophies of action or of perception. Instead, I will try to sort out a few things about the structure, mostly in an expository vein, mostly sketching some of the major themes. The chapter covers a lot of ground; in particular, action, perception, and epistemology. Most of the epistemological background in the book is in the chapter, preceding the perception stuff. The perception and action sections are supposed to present the theories of the book on those topics. It seems like neither is entirely satisfactory and a lot more could be said about all three parts.

The perception section bears most of the weight. Conceptually, action is identified with (I think that is right) language-exit moves while perception is identified with language-entry moves. Perception bears most of the weight because the model for perception is reused for action. The order of things is just, in a sense, reversed.

The model of perception is what Brandom calls the two-ply account. It consists of having an appropriate reliable differential responsive disposition and applying the appropriate concept. There is a two part structure here to reflect the interactions of the causal order of things and the rational order of things, with dispositions for the former and concepts for the latter. For action, instead of passively accepting a stimulus, the causal side of things is motivational. There is a further story to tell on the action side about practical reasoning and practical commitments, but this leans heavily on the established ideas of doxastic commitments, theoretical reasoning, and the two-ply model for perception.

Brandom is primarily concerned with materially good practical reasoning, like theoretical reasoning. This means that the inferences will in general be non-monotonic. Similarly to his views on material inference in the theoretical case, the practical case features multiple sorts of formal goodness. Whereas the different sorts of conditionals express different sorts of endorsements of inference on the theoretical side, the different sorts of oughts (instrumental, unconditional, etc.) express different sorts of endorsements of practical reasoning on the practical side. In a big way, the account of action is what you get if you take the structure for the theoretical side, i.e. inference, logic, perception and the rest, and change as few bits as possible to make it about action. The symmetry is both lovely and kind of creepy.