Defining normality--art or science?

The problem of defining normality, and in particular the establishment of reference ranges, is discussed. A so-called "objective" method of determining such ranges is described. It is suggested, however, that all currently used methods for defining "normal" ranges rely ultimately on a subjective choice. We propose the use of "How Often That Happens" (HOTH) graphs as an alternative to the conventional reference range. Such graphs show the frequency of results at least as extreme as any observed value. We have used such graphs, along with computer simulation models, in an attempt to describe how the frequency distribution of a given parameter changes as a population ages. Finally, an analogy is drawn between the states of "health", "illness" and "death", and the attractors which arise in non-linear dynamic systems. We conclude that the definition of normality is in fact more of an art than a science.