Abstract
This chapter discusses the limitations of approaches to modelling and handling evidential issues using hierarchical network representations. Such models of evidential inference rest on the compound proposition that real-world evidential inference usually or always consists of propositional ‘atoms’ that are linked together by nomological entities of some kind, entities that are often — but not always — called ‘generalisations’. These sorts of models or representations of evidential inference are referred to as ‘network-and-generalisation’ models of evidential inference. It is argued that for certain important problems, especially where these concern meaning and human understanding, these need to be complemented by other methods.