(1) I know that, at the present moment (t1), I have hands.(2) I might come at some future time (t2) to acquire goodevidence that I did not have hands at t1.(3) Were I to acquire such evidence at t2, then at t2 I couldnot know (even if I continued to believe) that I had hadhands at t1.(4) If S knows that p, and S continues to have all of hercurrent evidence for p, and continues to believe that pon the basis of this evidence, then S cannot lose herknowledge that p
by acquiring new evidence.Acquiring new evidence cannot, all by itself, make youmore ignorant.Different versions of this paradox are developed and discussed inHarman (1973) and Ginet (1980). The fact that these authors regarditasa
suggeststhatthereissomethingparadoxicalhere.Forinstance, Harman (1973, p. 148) admits that it seems paradoxicalto reject (4), but he does so anyway. And Pryor, like most otherepistemologists, would solve this paradox by rejecting (4), which isincompatible with the conjunction of (F) and (PK).Williamson (2000, p. 219) gives an apparent counter-exampleto (4). Here is his example: “I see one red and one black ball putinto an otherwise empty bag.
. . .
Now suppose that on the ﬁrst tenthousand draws a red ball is drawn each time, a contingency whichmy evidence does not rule out in advance, since its evidential proba-bility is non-zero. But when I have seen it happen, I will rationallycome to doubt e; I will falsely suspect that the ball only lookedblack by a trick of the light.” To see that this example doesn’t work to refute (4), consider that if I retain the evidence I had before thedraws, it should make no difference if the bag is transparent. But if the bag were transparent, then I would not rationally come to doubte. Rather, I would just be puzzled why my hand seemed always toreach for the red ball instead of the black one. Had I retained myevidence (by continuing to see the black ball), I would not rationallycome to doubt e, and so this is not a counterexample to (4).Of course, the plausibility of (4) does not make it immune fromrejection. But if we’re going to solve the paradox of defeasibility byrejecting (4), we should at least explain why (4) is so plausible, andso why there is a paradox here in the ﬁrst place. I’ll offer an alterna-