Barkley, I don't think it's quite right to call Keynes's approach 
"more sophisticated" than Knight's. Knight's frequentist approach and 
Keynes's subjectivist (or, in modern terms, Bayesian) approach are 
qualitatively different, each explicitly rejecting the other. Richard 
von Mises, perhaps the leading twentieth-century frequentist, wrote 
that subjectivists such as Keynes fail to recognize "that if we know 
nothing about a thing, we cannot say anything about its probability. 
 . . The peculiar approach of the subjectivists lies in the fact 
that they consider 'I presume that these cases are equally probable' 
to be equivalent to 'These cases are equally probable,' since, for 
them, probability is only a subjective notion." For the frequentists, 
such as Mises, Ronald Fisher, Jerzy Neyman, and Egon Pearson, the 
probability of a particular event is defined as the limit value of 
its relative frequency in a series of trials. In this understanding, 
probabilities can be defined only in cases in which repeated trials 
are feasible -- i.e., in situations where each event can be 
meaningfully compared to other events in the same class. Hence 
probabilities can only be defined ex post, as learned through 
experience, and cannot exist a priori. What Mises calls case 
probability -- i.e., uninsurable risk or Knightian uncertainty, is a 
situation in which probabilities, in the frequentist sense, do not exist.

I have the references in this short piece: 
http://web.missouri.edu/~kleinp/papers/09071.pdf. (I also note that 
Mises's use of the term "case probability" is misleading; what Mises 
really means is "case non-probability," or perhaps "case judgments 
without probabilities.")

Peter Klein