Making sense of outliers.

Phonetic inquiry, particularly of the kind which looks at large numbers of items, frequently asks what a 'typical' example of a phenomenon of speech is like. For example, in describing the plosives of a language, it is common to calculate average closure durations or voice onset times in some contexts. In this paper, the focus is on examples from natural conversation where there are untypical patterns, and where the item of interest is, in statistical terms, an outlier. Outliers are generally excluded from linguistic analysis, and may be treated as e.g. production errors. The paper shows instead that at least some outliers are in fact part of a meaningful practice, and an orderly method by which speakers can create meaning. The phenomenon more precisely is that of intensifying emphasis, which has been described for German. We consider its phonetic exponents in American English, which are both 'prosodic' and 'segmental'. We provide an account of function based on participants' own use and displayed understanding of the phenomenon, and argue for a structurally rich account of the phenomenon which includes aspects of turn construction and sequential organisation.