(Sorry for the breaks, but I'm writing this on-line, and I was about to be disconnected.)
If we follow Fetzer (24 Jan 1996) in thinking that sensations and perceptions are icons, then we have a concrete proposal, it seems, for how our cognitions involving the world are not subject to the sorts of permutation operations that make trouble for realism. In an earlier posting (28 Jan 1996), Paul Humphreys introduced into the discussion the idea that ``indexical icons'' could be used against the Goodman/Putnam line, and it would seem that if Fetzer is right that sensations and perceptions are icons, they would appear to be indexical icons, i.e., caused by the things that they represent. But Humphreys has balked at (or at least not endorsed) such a scenario at the end of his most recent posting.
We need to speculate about how minds use indices. Fetzer has suggested that a mind that uses indices must also be able to use icons - the mind must be able to recognise the resemblance between different instances of similar causal sequences. I'm not sure about this, but I think this suggestion assumes that the icons involved are mental (could an event in the world be an icon of another event in the world?). And they would also appear to be (or would be plausibly taken to be) indexical icons. This looks like a representational (or ``indirect'') realist view.
We need to hear/say more about our ability to recognise causal sequences: how is the recognition - and more importantly, the ensuing cognition (in Fetzer's sense) - non-propositional? Fetzer has claimed that although icons themselves are not assertive, cognitions involving icons tend to have an assertive character. Does this mean that these cognitions are propositional? If so, and if we want our ``causal cognitions'' to be non-propositional, we'll need a different picture. What makes indexical knowledge non-propositional?
One more question lurking in the background: indexical knowledge is, presumably, a species of perceptual knowledge, so how does our knowledge about causal relations differ from other perceptual knowledge with respect to our current set of concerns? Is there an important difference-with respect to the realism issue-between my knowledge that fire causes burns on my skin and my knowledge that there is fire in front of me? If there is not an important difference, then Humphreys' anti-anti-realist line of argument may be extendable to perceptual knowledge generally.