I understood sections 19 & 20 as saying:
"looks red" depends on the "concept of red". And "concept of red" depends on standard conditions and practice.
But what I want to be clear on is how he supports the first point, that "looks" depend on "concepts". As this is something that crossed my mind in my own writing (that someone might object "looks don't depend on concepts; therefore your points on public language don't apply").
I think the take-away from sect. 19 is that even seemingly "logically independent" fundamental concepts (apparently "characteristic of the empiricist tradition") really aren't. To repeat a previous comment,
I have found it useful to try to imagine myself in the position of a baby who "knows" almost nothing about anything and has to learn even things that only a few years later will have become so familiar as to seem to have been known all along, ie, to have been "given" just by virtue of being alive and having an intact sensory system.)Eg, to have the concept of a red triangle requires myriad related concepts: the general concepts of "object", "shape", and "color"; the specific concepts of "red", "triangular", and the mating of those; et al. That's the "holism" to which Sellars refers - the idea that concepts function in collections rather than in isolation.
I have to admit that sect. 20, being an imagined dialogue with those of a persuasion I don't quite understand, is somewhat confusing. The take-away for me is that Sellars' reply seems to be that whatever arguments may be mustered based on the ill-defined idea of "sense content", he has managed to (or at least ultimately will) develop a coherent argument with no reference to any such imagined entity. Which may explain why - as I noted in an earlier comment - he avoids what seems an obvious move, viz, to replace "sense content" with some idea of "neural state". It appears that he wants to make his argument as general as possible, in particular not to depend on reduction to any specific physiological underpinnings.
If one didn't have that objective, it seems to me that the resolution to the question of what is consistent among the three situations listed on p. 144 in the sect. 17 study guide is that all three could conceivably be associated with the same neural state. But the context for each situation could be different in that each could include a different collection of other relevant concepts, supporting information, etc. And that difference in context for the same neural state could explain the different levels of endorsement that define the three distinct situations.