Indeed, in "Formen des Scheiterns, Tadel, und Entschuldigung" et seq we get a little more on the topic, but at least one aspect of what Kern says in the named section is a little confusing. First she offers us three cases of privation that "together articulate the sense which the claim that being fallible is a formal characteristic of these capabilities" [why did I translate this when I have no intention of translating what follows? I have an idea, actually!]:
Jemand hat den sinnlichen Eindruck, daß p, ohne wahrzunehmen, daß p. Jemand hört einen anderen sagen, daß p, ohen von einem anderen zu erfarhen, daß p.
Jemand glaubt, wahrzunehmen, daß p, ohne wahrzunehmen, daß p. [likewise for hearsay]
Jemand nimmt wahr, daß p, ohne zu glauben und also zu wissen, daß p. [likewise for hearsay]
The last example is supposed to explain situations like this (this is Kern's example): the parents of a child don't believe it when their child says there's a fire in the house, because the child is constantly saying just that, falsely; this time he is saying it truly, and so there is an opportunity for the parents to gain knowledge that there is a fire, which they do not take because they are in reflectively unfavorable circumstances—that is, the circumstances in this case are favorable, but there is reason for the parents to think that they are not favorable.
This is confusing, however. In what sense, in this boy-who-cried-wolf scenario, are the circumstances actually favorable—at least if we want to maintain the claim that the person in the barn-facade example can't know that he is looking at a barn? (In fact the barn-facade example doesn't seem to fit easily into the above tripartition; in it one perceives a barn and believes it to be a barn, but doesn't know it to be a barn.) This is all the stranger because Kern, to dramatize the reflective unfavorability of the circumstances, introduces an aunt into the family, who, because she only visits every two years, does not know of the child's lately acquired habit of crying "fire", and comes to know, on hearing this one true "fire" report, that there is a fire in the house. But how, one might well be inclined to ask, does this situation essentially differ from that of the traveler into barn facade country, who, because he doesn't read guidebooks, does not know of the locals' peculiar habits, yet cannot come to know, on looking at the one real barn, that it is a barn? Suppose it's the only structure he actually looks at at all in his trip (eyes on the road, y'know). None of the explanation for why the barn-looker doesn't have knowledge involved reflection.
A slight shift: Adrian Haddock has an article in a recent (Mar '11) issue of Philosophical Explorations (and the subsequent issue has an annoying article by some fool attempting—badly—to trespass on my dissertation; how aggravating) on disjuctivism, part of a series including a response by Burge to Mcdowell's response to Burge. Haddock wants to motivate a disjunctive conception of perceiving alongside (I take it alongside, anyway) that of experience; some perceivings are mere perceivings (as in the barn-facade case, or in reflectively unfavorable circumstances); some are perceivings of such-and-such that "are cases of being in a position to know that one perceives such-and-such", where this "being in a position" excludes non- and reflectively unfavorable circumstances. He mentions the barn-facade style of case but says he does "not want to place much weight on it", partly because he is "not sure it merits much weight" and partly because he thinks there's a much more worrying problem. But the way he motivates that more worrying problem leaves me utterly unmoved.
It is an adaptation of an argument of Williamson's that seems in essence to date at least to Russell; it hinges on endorsing this principle: "for any times t and t + 1, where t and t + 1 are any two times spaced only fractionally—say, one millisecond&mash;apart, if at t one knows that something is at the case, then at t + 1 this very thing is the case; e.g. if at t I know that I see that your sweater is brown, then at t + 1 I see that your sweater is brown." (But isn't this principle obviously false? Surely I can know something that ceases to be the case within a millisecond. Suppose I learn (it doesn't matter how) that there is a pen in my office. And sometime later someone detonates a stick of dynamite in my office, destroying the pen within a millisecond (at t the shock wave is still approaching it; by t + 1 it's up in smoke). Why should we think that at t I didn't still know what was after all still the case at t, that there is a pen in my office?) Haddock thinks that it would be difficult to deny this principle. Part of the argument is:
Now it seems that, if in one situation something is the case and one believes that it is the case, but in a very close situation it is not the case and yet one still believes that it is the case, then the claim that, in the former situation, one believes that it is the case because it is the case is in jeopardy. … What is it for two sitations to count as being "very close"? Well, it seems to be a fact about our capacity for altering confidence-levels that we cannot, across a fractionally small space of time (say, a single millisecond) shift from a state of believing that something is the case with the confidence required for knowing that it is the case, to a state of not believing that it is the case. And here by "our capacity" I mean the capacity of human beings like us … We might cast this fact into hypothetical form by saying that, if at one time one knows (and so believes) that something is the case, then at a fractionally later time (say, one millisecond later) one still believes that it is the case … Given this, it seems that two situations temporally spaced only fractionally (say, one millisecond) apart will always count, for human beings, as "very close", no matter how much the situations may differ in other respects.
Assuredly, they will count as temporally very close, while for those whose belief-revision processes are yet laggier situations a minute or an hour apart will count as temporally very close. But why on earth would this fact make one think that the claim about jeopardy holds? Suppose our abilities to revise our beliefs were indeed quite laggy, so that the minimum lifespan of a perceptually formed belief is not around one millisecond but more like an hour. Belief formation is still swift; the beliefs thus formed just stick around for a long time. Then we can ask two questions: (a) why was this belief formed, why does it have the content it has? (b) why does he (still) have it? The answer to (b) will be something like "once it's formed, it just sticks around for a long time"; the answer to (a) can still be "he believes it because it was the case when he formed it" or even "he formed it because it was the case". The change in time scale only makes clear that these different questions are both applicable and are different; at the shorter scale it still applies. That I am unable to register changes that occur within a certain time scale doesn't mean that when I do register something I don't register it because it is the case; we have different explanations for the formation and the persistence of the registration. I really can't see how this is supposed to be convincing, thus why we should bother with the anti-luminosity argument as Haddock presents it, especially since the barn-facade case <em>anyway</em> motivates his position (and resembles, if not recapitulates, situations we actually encounter in real life) and the anti-luminosity argument trades on vagueness in a to-me suspicious way.