With regard to change in inflection, historical linguistics fundamentally relies on the concept of morphological analogy, which is held responsible for nearly all change not attributable to phonological factors. Despite its central importance, how morphological analogy operates has never been established. Two different opinions are held in contemporary linguistics. The first position assumes that morphological analogy modifies inherited inflectional forms, making them more similar to other inflectional forms. According to the second position, in the course of morphological analogy, inherited inflectional forms are not merely modified but rather are replaced by forms created entirely anew on a model pattern already present in the grammar. This research report tries to establish what kind of data may constitute the evidence sufficient to differentiate between the two views. It argues that all relevant data point to whole-word replacement as the only mechanism of analogical change in inflection.