From coda to coda, from town across town: The rise of coda approximant r in Dutch

The realisation of /r/ in Dutch is notoriously variable across speakers and across linguistic contexts, with recent studies distinguishing around twenty variants in Standard Dutch alone (Smakman 2006; Sebregts 2014). The variation in place of articulation ranges from alveolar to uvular, and manner of articulation from trills and fricatives to approximants and vowels. One important r-variant in Dutch is a coda approximant [ɻ], itself variable articulatorily across speakers: as shown by ultrasound studies, it can be either retroflex or bunched pre-velar (Scobbie and Sebregts 2010). This variant appears to be on the rise in a typical ‘change from above’ fashion: it is highly prestigious, and speakers are relatively aware of the variant’s use (Van Bezooijen 2005). The restriction to coda position for this variant leads to relatively complex systems of r-allophony within speakers, who mostly combine the approximant with alveolar taps or uvular trills/fricatives.

Linear mixed-effects regression analysis of a large-scale corpus of colloquial Standard Dutch from various urban centres in the Netherlands (analysed auditorily and acoustically) shows that use of the retroflex/bunched approximant in coda is a rapid change-in-progress in cities especially in the western Netherlands, with young female speakers at the forefront: in Leiden, Rotterdam, and The Hague, there are significant effects of age, and in the latter two cities also of sex. In other cities, such as Amsterdam and Utrecht, there are similar trends, but the effects are not large enough to reach significance.

There is also a difference between the two sets of cities in the linguistic distribution of coda approximant r. The variant is most frequent in ‘true’ coda positions, i.e. word-finally and in clusters before a final coronal consonant (e.g. schaar /sxar/ ‘scissors’ and kaars /kars/ ‘candle’). In clusters with non-coronal consonants, /r/ is in a ‘potential’ coda position in Dutch: there is an optional process of schwa-insertion in liquid-obstruent clusters (Booij 1995), creating an additional syllable and moving /r/ into an onset (e.g. kerk /kɛrk/ [kɛ.rək] ‘church’). Schwa-insertion and use of the retroflex/bunched approximant only very rarely co-occur: items such as kerk are either produced with schwa-insertion and an onset variant of /r/ ([kɛ.ɾək] or [kɛ.ʁək]) or with a coda approximant but without schwa-insertion ([kɛɻk]). In those cities that show the clearest age and sex effects in use of the retroflex/bunched approximant, its use in the schwa-insertion context (and concomitant blocking of schwa-insertion) is also more frequent. This effect appears independent from the token frequency of the variant: while the overall token frequency of [ɻ] in Rotterdam and Utrecht is similar, the variant has made inroads into the schwa-insertion context in the former but not the latter.

The rise of the retroflex/bunched approximant in Dutch, then, is along two routes: linguistically, it is spreading from the true coda context to the ‘potential’ coda context, blocking the process of schwa-insertion as it does so. Socially, it is young and female speakers who are leading the change, as expected in a change from above.