Let R(X, B) denote the class of probability functions that are defined on a
lgebra X and that represent rationally permissible degrees of certainty for
a person whose total relevant background evidence is B. This paper is conc
erned with characterizing R(X, B) for the case in which X is an algebra of
propositions involving two properties and B is empty. It proposes necessary
conditions for a probability function to be in R(X, B), some of which invo
lve the notion of statistical dependence. The class of probability function
s that satisfy these conditions, here denoted P_I, includes a class that Ca
rnap once proposed for the same situation. Probability functions in P_I vio
late Carnaps axiom of analogy but, it is argued, that axiom should be rejec
ted. A derivation of Carnaps model by Hesse has limitations that are not pr
esent in the derivation of P_I given here. Various alternative probability
models are considered and rejected.