The criticism, made vivid by the tale of Bayesian Peter, is that these ‘ampliative’ rules are vulnerable to a Dutch Book. Adopting any such rule would lead us to acknowledge as fair a system of bets that foreseeably leads to certain loss. Any rule of this type for analogical reasoning appears to be vulnerable to van Fraassen’s objection. There appear to be at least three routes to avoiding these difficulties and finding a role for analogical arguments within Bayesian epistemology. First, ther