Dutch /r/ variation and the representation of complexity

Rhotics, or /r/ sounds, show a great deal of variation across the world’s languages. The situation in Dutch is striking in that virtually all known rhotics occur as allophones of its one rhotic phoneme. A feature representation of /r/ in Dutch in terms of Element Theory (Kaye et al 1985, Harris and Lindsey 1995) can take one of two directions. As has been suggested for English (e.g. Harris 1994) and Chinese (Kaye 2000), /r/ can be thought of as almost empty, i.e. consisting of only one element (e.g. the ‘low’ element |A| or the neutral element); this would leave the phonetic interpretation relatively open. Alternatively, /r/ can be said to consist of a relatively large number of elements, as has been suggested by Scheer (1999) and, specifically for Dutch, Van der Torre (2003), which may in turn give rise to many possible variants.
I will show that the second option, as implemented by Van der Torre (2003), is able to account for the large number of /r/ variants found in Dutch. A survey of Standard Dutch in urban communities in the Netherlands and Flanders will serve to illustrate how the suppression of one or several of the elements that make up /r/ leads to the variety of realisations in different contexts. These variants, ranging from fricatives to vowels, should therefore be seen as constituting reductions of this highly complex sound. Note that ‘reduction’ in this sense does not denote a step upward on the sonority scale, the traditional notion of lenition. Instead, it is on a par with what Schiller (1999) calls the reduction variants of German /R/: all forms that are (phonetically) simpler than the complex uvular trill.
In addition to being well-suited for the description of /r/ variation in Dutch, the approach to the representation of /r/ in Dutch outlined above clears up a persistent issue within Element Theory. Whereas for the representation of vowels within Element Theory internal complexity is consistently the inverse of cross-linguistic frequency (a ‘markedness’ effect), this is not the case with consonants. Instead, complexity in consonants is linked to the sonority hierarchy: the least sonorous sounds have the most complex representations. This leads to a situation in which the most frequently occurring consonants, oral stops, are represented as most complex.
In Van der Torre (2003), however, the sonorant consonants have the most complex representation; both obstruents and vowels, though opposites in terms of sonority, are representationally simpler. Liquids in particular are viewed as at the top of this complexity scale, as they consist of four elements (including two resonance elements) each. This is more in accordance with cross-linguistic frequency of phoneme classes. Under this approach, phonetic complexity is equated with phonological complexity, which, as I will show, serves to explain the variation found with Dutch /r/.