This paper describes a program called Hob that uses analogical mappings across narratives to drive an experimental conversational system. Analogical mappings are used to drive internal reasoning processes and supply response templates. The knowledge base is written in simple English, and consists of three parts: a dictionary, a collection of facts and heuristics, and a collection of stories that notionally correspond to an experience base. Internally, knowledge is stored and used in the form of near-canonical English parse trees. Thus the schema is not hardwired, but is derived from the knowledge base as interpreted under the rules and implications of English grammar. An experimental "sellbot" application is described, and example runs are presented.