If the warrantWarrant of an argument is defeasible, do the premisesPremise support the conclusionConclusion with PascalianProbabilityPascalian probability? We question this assumption. We find each of the five “standard” interpretations of PascalianProbabilityPascalian probability wanting. Furthermore, the PascalianProbabilityPascalian complementational principle, if Pr(C/P) = .r, Pr(~C/P) = 1 − .r, need not hold. The multiplicative principleMultiplicative principle indicates that if an expert vouches for a conjunction of statements, his vouching is weaker than his vouching for either statement by itself. But why? By contrast, epistemic probabilityEpistemic probability interprets argument strength as degree of resistance to falsification by relevant variablesRelevant variable. Given a warrantWarrant, “From Px To infer Cx” assume some instance, “Ra” rebuts the inference from “Pa” to “Ca.” “Rx” is a relevant variableRelevant variable. If some instance “Ra” does not rebut the inference from “Pa” to “Ca” the warrantWarrant resists that instance. The more instances resisted, the stronger the warrant. Following CohenCohen, L. Jonathan, we may construct a canonical testCanonical test of warrant strength through resistance to relevant variablesRelevant variable. The number of levels of the test passed before a counterexample appears the greater the strength of the warrantWarrant. We address whether epistemic probabilityEpistemic probability gives a proper account of connection adequacy by examining in subsequent chapters the distinct types of defeasible warrantsDefeasible warrantEpistemic probability.