Rhotics, or r-sounds, are known to display a large amount of variation, both cross-linguistically and within particular languages. Dutch is an example of such a language: even within what is generally regarded as the standard varie-ty, around 20 different variants can be distinguished, and the variation among them is governed by both linguistic and extra-linguistic factors. The present study has both empirical and theoretical aims. Its main empirical aim is to catalogue Dutch r-variation, based on the collection and analysis of a corpus of acoustic speech data from over 400 speakers (~20.000 tokens) in ten cities in the Netherlands and Flanders. In addition, it presents a detailed articulatory (ultrasound) study of an innovative coda approximant variant. The theoreti-cal contribution of the thesis lies in the development of a model of progressive sound change to account for the origins, development and current status of Dutch r-variation. In doing so, it tries to unite the interests of sociophoneticians and phonologists in the study of r-variation: while the former may seek to un-tangle the geographical, social and linguistic factors involved, the latter might attempt to explain what unites the many variants, in order to define r as a category. This thesis shows the relevance of detailed sociophonetic analysis for answering phonological questions, and argues that the unity among r-variants is established by tracing the diachronic links between them.