In a number of publications, John Anderson and his colleagues have developed the notion of Structural Analogy — the assumption that structural principles generalise across levels of language — as a meta-theoretical principle of linguistics. In this paper some recent works which appear to criticise the notion are considered, and it is concluded that they fail to invalidate it. It is suggested, however, that some work by Anderson and his colleagues in Dependency Phonology appears to misuse the notion to mask differing content for the same structural labels. It is suggested that the notion of Structural Analogy may not be a scientific notion at all, and should not be used to make predictions about linguistic structures.