With regard to the subject of my last post, a post has been put up at Studi Galileiani on verisimilitude in history. Now, I first need to say that there are a lot of issues related to this subject I haven't thought through (hence my term 'naked' in the previous post); and the conclusions I have come to are inevitably idiosyncratic (they always are). So none of this is intended as more than a crude set of first thoughts. My thoughts on the post:
* I'm inclined to think that the 'theory-laden' nature of our evidences is not actually relevant to the issue. It only would become so if we thought that the ability to approximate reality actually depended on the ability to separate observation statements and theory statements. But there is no reason to assume this unless we are positivists. In fact, I think the 'theory-laden' nature of our evidences is a boon to a non-positivist realism, because it is precisely what makes it possible for us to engage in any inquiry at all (I follow Whewell to that extent). It is what makes them intelligible evidences rather than unintelligible brute facts.
* One effect of going to approximation, it seems to me, is that it does away with any demand for our histories to imply "a unique historical reality against which to measure the fit our histories." Measurement against a unique historical reality is not what being approximately true involves; in other words: you don't need to have in hand a unique historical reality in order to determine that a history is approximately true. Just as you don't need to know that the earth is a perfect sphere in order to say that "The earth is a perfect sphere" is approximately true (which is a good thing, since the earth is not a perfect sphere), so you don't need to know a unique historical reality in order to say that such-and-such history is approximately true. You just need to know some relevant reality; and there is no non-positivist reason to think our theory-laden traces of the past do not qualify as relevant reality. (A sense of the term 'reality' in which an archive does not count as real is a bizarre sense that may simply be discounted; such a sense would be irrelevant to anything and not what realists mean when they talk about reality. And the mere fact of conceptualizing it as an archive makes it relevant. Skeptical issues can arise about the ways in which it is relevant in any particular case; but once admit it an archive, you admit it is a reality relevant to building approximately true histories around.)
* If we take approximate truth seriously, an appeal to coherence doesn't require us to appeal to a coherence account of truth, but only to a coherence account of approximate truth; it is true that mutually exclusive accounts may both cohere in the way such an account requires, but that's not a problem for approximate truth. In fact, it is precisely what one would expect. Nor is it necessary in every case to rank the accounts relative to each other and say, this account is a better approximation than that one; such a ranking would be linear, but we are dealing with accounts, not single propositions, and there is no reason to think that there would always be a straightforward way in which this account is simply a better approximation than that one. Accounts are complicated things; things may approximate to reality in qualitatively different ways, i.e., while the approximation might be compared, there may be no common measure because they approximate in different ways. At least, it is arbitrary simply to assume otherwise at the outset. So the question "Which has the greater verisimilitude?" is only a bugbear for the historical realist if the historical realist is assuming that all historical accounts would be approximations of exactly the same thing in exactly the same way, so that each approximation could be ranked as more or less approximately true according to a single measure than every other approximation. But this contains a is a lot of strong modifiers a historical realist doesn't have to assume.
* History is narrative discourse, and a lot like literature, amen, amen, I say to you, amen. But it is entirely possible to be a realist and hold this.
I haven't looked at Dummettian semantic anti-realism or anti-representationalism here; I suspect there are stronger issues there, although I'm too fuzzy on both to say anything constructive. But I don't see that there's anything for the historical realist to fear in bias, underdetermination, or the theory-laden character of our evidences. They only really are a problem when we are trying to find a unique historical reality rather than approximation to it; they only arise under the latter case if we assume that to recognize something as approximately true requires a unique reference point, i.e., they only arise if one assumes that approximate truth requires both the approximation and the unique historical reality. I see no reason to think this is true.